texonom/texonom-md · Datasets at Hugging Face

title stringlengths 1 544 ⌀	parent stringlengths 0 57 ⌀	created stringlengths 11 12 ⌀	editor stringclasses 1 value	creator stringclasses 4 values	edited stringlengths 11 12 ⌀	refs stringlengths 0 536 ⌀	text stringlengths 1 26k	id stringlengths 32 32
Claude Shannon	Computer Scientists	May 9, 2023	Alan Jo	Alan Jo	Jun 25, 2023	[Alan Turing](https://texonom.com/alan-turing-df61b5084e544ac58577d289983e0d28)	구글 학술검색 기준으로 세상에서 4번째로 많이 인용된 논문인 <정보통신의 수학적 이론>이라는 논문을 발표 [Entropy](https://texonom.com/entropy-35fdf9bb7fdd42edbd4277147ca0ec38) 1950~1986 연 평균 복리수익율이 28% 로 세계최고수준 [N gram Language Model](https://texonom.com/n-gram-language-model-e600ddf01e2645e1b1d76604454bc3c9) [One Time Pad](https://texonom.com/one-time-pad-0fd4b702a7854dca8263035014fa3914)	bc27d0c33de4438cb326957a169ed36a
Edsger Dijkstra	Computer Scientists	Aug 9, 2023	Alan Jo	Alan Jo	Aug 9, 2023		> [Edsger Dijkstra's One-Day Workweek - Cal Newport](https://calnewport.com/edsger-dijkstras-one-day-workweek)	569bf90716e64b468e888f51099c8beb
Grace Hopper	Computer Scientists	Sep 7, 2020	Alan Jo	Alan Jo	Jun 25, 2023		최초로 컴파일러를 만들고 처음으로 디버깅(?)을 한 그레이스 호퍼도 여성 1947년 호퍼는 마크2에서 프로그래밍 작업을 하고 있었는데, 그 날 따라 천공카드 입력이 잘 안되었다. 결국, 릴레이 안에 죽어있는 나방 때문에 컴퓨터가 제대로 동작하지 않는다는 사실을 알게 된 그리고 그 나방을 작업 노트에 붙이면서 최초의 컴퓨터 버그라고 메모를 남긴다.	1b108f44e96a4899b18c1a6821869f48
Margaret *Hamilton*	Computer Scientists	Sep 7, 2020	Alan Jo	Alan Jo	Jun 25, 2023		### software engineering 이라는 용어를 처음 사용했다 대학에서 수학을 공부한 그녀는 남편의 박사 학위 공부를 지원하기 위해 MIT에서 프로그래머로 일을 시작했다. 당시 그녀는 제대로된 프로그래밍 교육도 받지 못하고 바로 날씨를 예측하는 프로그램 개발에 참여 그 후, 시스템 프로그래밍 전문가가 되어 아폴로11 프로젝트에 참가하게 된다. 당시 아폴로11 프로젝트에는 SW개발을 위한 예산과 일정이 포함되어 있지 않았고, 요구 사항에도 소프트웨어 대한 언급은 없었다 > [F/OSS Story - 페이지 5 - 만화로 나누는 자유/오픈소스 소프트웨어 이야기](https://joonecomics.wordpress.com/category/foss-story/page/5/)	ea62b054dc0a47d291b80dc976b5f8d8
Shumuel Winograd	Computer Scientists	Jun 25, 2023	Alan Jo	Alan Jo	Jun 25, 2023			5f38451bbad04b6da44260f93ca4eb43
Von Neumann	Computer Scientists	Apr 15, 2023	Alan Jo	Alan Jo	Jun 25, 2023			ef79cb1b1e9045d28caced75663e14f7
Alan Turing History	Alan Turing Notion	May 9, 2023	Alan Jo	Alan Jo	May 9, 2023		케임브릿지 대학교 입학 프린스턴 양자역학 논문에 영감을 받음 그 논문이 존 폰 노이만이 쓴 것 힐베르트가 기계가 만능이다라는 주장을 했고 세계수학학계가 암울했는데 학교에서 기계가 만능이 아니다라는 의미를 가지는 [불완전성 정리](https://texonom.com/e158a278de5b4a7fa97b509cbd2dc0bd) 를 읽고 연산을 수행하는 자동화된 기계라는 분야에 흥미를 가지고 공부하게 된다 괴델의 불완전성 원리의 간단함에 새로 자신만의 방식으로 증명 그 과정에서 universial machine이라는 단어 나옴 이런 방식으로 람다대수를 만들어서 불완전성 원리 증명한 알론소 처치를 따라 프린스턴으로 떠남 만들어냄 22살에 수학박사되고 논문 엄청 쓴 존 폰 노이만과 교류 맨하탄 프로젝트에 존 폰 노이만이 참여 앨런 튜링은 암호 얀그서 에드박 참여 논문 전쟁 끝나고 냄 근데 중단됨 못알아봐서 이후에 몇년뒤 구현되었는데 당시 세계젤빠른 컴퓨터 폰노이만이 edvac 보고서 미국서 썻는데 튜링머신 아이디어 참고했는지 확실하지 않다 이후 영국 현실에 막혀 맨체스터 대학에 가서 인공지능 공부 수학과 생물학 공부하며 논문 쓴다 과연 기계는 생각을 할수 있을까라는 질문으로 논문 시작 철학적 논문can machine think? Imitation Game 여기서 나오는 말 , 강화학습 개념 소개 모든 지성의 영역에서 인간과 기계의 경쟁을 예측 이후 자연에서 발생하는 패턴이 생기는 과정을 수학적으로 서술 호르몬 화학치료중 사과먹고 자살 폰 이만 구조 이론적 창시자 AI의 선두자이자 튜링 테스트의 어원 애니그마 개발자 어린시절 튜링은 친구를 되살리고 싶은 마음에 양자역학과 수학에 심취했다 엘런 튜링의 기계에 대한 증명은 괴델의 불확정성 원리에 기반한다 2차대전 에니그마 독일 이 uboat (undersea)에서 사용한 암호화기술 - 1943 - 앨런 튜링 에니악도 수소폭탄 시뮬레이션 위해 만들어진 것 - 1946 - 미국 폰 노이만이 edvac 보고서로 폰 노이만 구조 정립 (1936 앨런튜링이 제안한 논문과 거의 비슷 - 하지만 튜링 언급 없다) 엘란 튜링이 보니 불완전해보이는 보고서 → 영국에서 만들어주세요 해서 더 대단한 Proposed Electronic Calculator라는 논문을 낸다 Automatic Computing Engine 으로 이름짓는다 edvac을 넘어 서브루틴 가능하게 구조 설계했지만 너무 복잡하다고 영국에서 불가능한거 아니냐고 못알아먹음 ace프로젝트 중단됨 3년뒤 pilot ace라는 계속기로 제작되서 세계에서 가장 빠른 컴퓨터로 기록된다 그 후 궤델 소개해준 맥스뉴먼 따라 맨체스터 대학 가서 연구 모두 초월한 인공지능을 만들어야겠다 해서 수학적 생물학 공부 시작 1950년에 - computing machinery and intelligence를 만든다 The imitation game이라는 유명한 단어가 나온다 → turing test 라고 불리며 매년 세계에서 열리고 있다 2014년에 유진 구스트만이 최초로 튜링 테스트를 통과한다 이 논문 마지막으로 한문장으로 정리 - 철학적 논문 이후에 직접 turochamp를 개발하게 되지만 진짜 실행에는 실패 - 알고리즘 복잡 2012년도가 되서야 만든다 - 결과는 좋지 못했지만 morphogenesis라는 개념 자연에서 발생하는 패턴이 생성되는 과정을 수학적으로 설명하는 논문도 내게 된다 - 생물학 논문 아놀드 머레이라는 남자와 연인관계했지만 동성애 영국 불법이라서 구금된다 2013년에 영국은 동성애결혼이 합법화된다 호르몬치료를 받으며 슬픔을 못이겨 청산가리를 만들어 사과에 넣고 자살한다 이후 영국은 에니그마를 다른 국가에 보안 확실하다고 팔고 자기들만 해독해서 그 정보 사용한다 영국은 이를 감추기 위해, 동성애자 앨런 튜링을 감춘다 the ultra secret에서 튜링의 업적이 밝혀진다 영국과 애니그마 2009년에 앨런튜링 사면운동이 일어난다 - 2010년에 스티븐 호킹이 가세하고 10대 인물에 엘런 튜링이 들어가고 엘리자베스 2세 여왕의 특별사면권으로 공식 사면된다 컴퓨터계의 노벨상인 튜링상이 생긴다 > [앨런 튜링 - 현대 컴퓨터 과학의 아버지 (feat. 세계 2차 대전) [상]](https://www.youtube.com/watch?v=QasRpCP8qsI) > [2. 앨런 튜링과 폰노이만](https://joone.net/2016/11/15/%ec%95%a8%eb%9f%b0-%ed%8a%9c%eb%a7%81%ea%b3%bc-%ed%8f%b0%eb%85%b8%ec%9d%b4%eb%a7%8c/)	b394066bcd2a49fb9e048836bbf1a212
Automatic Computing Engine	Alan Turing Notion	Mar 2, 2021	Alan Jo	Alan Jo	May 9, 2023		처음으로 메모리 내장 이용 cpu memory 입력장치 출력장치 설계	a61d7bc038fe4e58a80989d1c3e6facf
OS Function	OS Notion	May 19, 2021	Alan Jo	Alan Jo	Jan 18, 2023		### OS Functions \|Title\| \|:-:\| \|[Processor Scheduling](https://texonom.com/processor-scheduling-924d2fa78737494d82e74595de81219f)\| \|[Memory Management](https://texonom.com/memory-management-dbf37c509eaf46deb2df8ccd086e386c)\| \|[Disk Management](https://texonom.com/disk-management-f82047a4bc5a4bfb9960448cd13ac476)\|	cc6a2b8fffb94340b51bc3f4a9c38862
OS History	OS Notion	Nov 5, 2019	Alan Jo	Seong-lae Cho	Jun 21, 2022		초기 컴퓨터들은 [Punch Card](https://texonom.com/punch-card-dcb71893cc074afb9ec6865b43eaaa61) 를 입,출력에 사용 물리적으로 컴퓨터 메모리에 로드해서 실행시키고 실행이 완료되면 메모리 덤프를 해서 출력은 얻음 [TTY](https://texonom.com/tty-d3c35ff3460343359b453c88a3bb896f) [Punch Card](https://texonom.com/punch-card-dcb71893cc074afb9ec6865b43eaaa61) ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2Fc55655c1-5711-44cc-be47-ee9da8f95035%2FUntitled.png?table=block&id=2248fd3f-a36c-4f4b-a903-2b3d3c4d3548&cache=v2) 이 과정을 사람이 직접 반복하다가 프로그램에 의해 자동화 하기 시작한것이 Operating System 의 시작 > unix: bsd (now free bsd), system V: bsd: for 학문: system V: ex. 솔라리스 >> init in bsd: getty >> init in system V: inittab >> those are psd 1 > > UNIX → GNU → Linux > → POSIX → BSD → MACOS > ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F8666dfa6-0c1f-4d3f-be0c-245ae45e6ee4%2FScreenshot_20200509-172659_Chrome.jpg?table=block&id=c42adbe0-47e2-4bd2-8ae4-33e1627aba09&cache=v2) > [Introduction](https://mug896.github.io/bash-shell/tty.html)	0f04c35c56144da9a9646d935a4a69fd
OS Structure	OS Notion	May 19, 2021	Alan Jo	Alan Jo	Apr 3, 2023	[Software Architecture](https://texonom.com/software-architecture-5d3c80f6ffe44d60b10c870903c2b703)	### OS Structures \|Title\| \|:-:\| \|[CLI](https://texonom.com/cli-3b46069e77904f7b8ce85faff6252cfc)\| \|[OS Kernel](https://texonom.com/os-kernel-631bce6131a64accb3a482df622ff957)\| \|[Disk System](https://texonom.com/disk-system-104aed644fd54c31a9938e4f7b3bae6f)\| \|[File System](https://texonom.com/file-system-7a340052fcaf48169a5c6da2610aa5cf)\| \|[I/O Systems](https://texonom.com/io-systems-1fe831bd09ae43909cccb33056a1dfc7)\| ### OS Components \|Title\| \|:-:\| \|[GUI](https://texonom.com/gui-936270e7b4af43e3929f69fcb6e1e2e4)\|	a66cf2a9ea8f4841b64ca65fbec8b743
UNIX	OS Notion	Mar 18, 2021	Alan Jo	Alan Jo	Apr 25, 2023		## UNiplexed Information Computing System ### Unix Notion \|Title\| \|:-:\| \|[POSIX](https://texonom.com/posix-6fd6ba0425f145fca4dea48a6dee1ce7)\| \|[Unix Philosophy](https://texonom.com/unix-philosophy-8f96f33ed93b4ea5a533303869abfa92)\| \|[UNIX Structure](https://texonom.com/unix-structure-f892f7c631854084a7a93cf8db918cc3)\| \|[Unix-like](https://texonom.com/unix-like-578ba7dbb07b44b3a2d3895ca103c5ec)\| \|[UNIX History](https://texonom.com/unix-history-34c9169587714b508aad541945d7571d)\|	233213b96425492f925a8277650c5b28
Disk Management	OS Functions	Nov 5, 2019	Alan Jo	Seong-lae Cho	Jan 18, 2023	[File System](https://texonom.com/file-system-7a340052fcaf48169a5c6da2610aa5cf)	### Need - Ram and disk price difference is still 10^2 - Memory is volatile - Disks as swap space for virtual memory system ### Disk Management Notion \|Title\| \|:-:\| \|[Disk layout](https://texonom.com/disk-layout-5091a147a01a4ff2986c9d80e6aad326)\| \|[Disk Allocation methods](https://texonom.com/disk-allocation-methods-4a49727bed65460cb4db6426038b97b4)\| \|[Disk caching](https://texonom.com/disk-caching-f0d66b9112c24fbd9b6bb7f115da6af0)\| \|[Anatomy of hard-disk](https://texonom.com/anatomy-of-hard-disk-e4eca2717d6145a189df99f477adf720)\| \|[Disk Head Scheduling](https://texonom.com/disk-head-scheduling-e6e60cca761147059a1bf7717616c398)\| \|[Disk management](https://texonom.com/disk-management-2db57b8f7837407088208b308e51bbec)\| \|[RAID](https://texonom.com/raid-4b50ffdd255d430ca76f393e8d1ebe25)\| \|[Tertiary storage devices](https://texonom.com/tertiary-storage-devices-def6411e9adf4220811b65706895e6e3)\| \|[Disk Performance](https://texonom.com/disk-performance-28e569d80cbe4c3b9e68857ac92550ea)\| > primary - secondary - tertiary	f82047a4bc5a4bfb9960448cd13ac476
Memory Management	OS Functions	Nov 5, 2019	Alan Jo	Seong-lae Cho	Jan 18, 2023		### Memory Management Notion \|Title\| \|:-:\| \|[Memory Management History](https://texonom.com/memory-management-history-baac71af6c7c4281894e15349bde6777)\| \|[Memory Management Method](https://texonom.com/memory-management-method-040223391bea4c148bcfee4f4064572f)\| \|[Memory Address Changing](https://texonom.com/memory-address-changing-53916e26496843d6bf5a26e6185ebe8d)\| ### Memory Management Usage \|Title\| \|:-:\| \|[Garbage Collector](https://texonom.com/garbage-collector-8af9f92e3ca842ee8ddd638e25647cd7)\| \|[Memory Leak](https://texonom.com/memory-leak-16c4f840b59e4d45a04770b03e582229)\| ### reference bit > [Untitled](https://asfirstalways.tistory.com/140) ### associative register > [연관기억장치, 복수 모듈 기억장치](https://devkim93.tistory.com/entry/%EC%97%B0%EA%B4%80%EA%B8%B0%EC%96%B5%EC%9E%A5%EC%B9%98-%EB%B3%B5%EC%88%98-%EB%AA%A8%EB%93%88-%EA%B8%B0%EC%96%B5%EC%9E%A5%EC%B9%98)	dbf37c509eaf46deb2df8ccd086e386c
Processor Scheduling	OS Functions	Nov 5, 2019	Alan Jo	Seong-lae Cho	Aug 10, 2023		### Scheduling is small issue in OS but hard to optimize ### Processor Scheduling Notion \|Title\| \|:-:\| \|[Processor Scheduler](https://texonom.com/processor-scheduler-8c124741c1e84d26b0b0a8ad599337f4)\| \|[Process](https://texonom.com/process-ffd68f9e2b2648759b96047d48d17778)\| \|[Interrupt](https://texonom.com/interrupt-d26d34c54d5447a6888ad85089184fb3)\| \|[Thread](https://texonom.com/thread-d704a229d3b94619963c61bb742b822b)\| ### Processor Scheduling Usages \|Title\| \|:-:\| \|[Processor Scheduling Algorithm](https://texonom.com/processor-scheduling-algorithm-305e26d3aca84cdd9e5c980e91642a8f)\| \|[Processor Scheduling Method](https://texonom.com/processor-scheduling-method-c5c243a128b74b2183508330aec1090c)\| \|[Processor Scheduling Evaluation](https://texonom.com/processor-scheduling-evaluation-b897d00ffba948e6922c062d95693455)\| > [[OS] 5-2. CPU Scheduling II](https://asfirstalways.tistory.com/124)	924d2fa78737494d82e74595de81219f
Anatomy of hard-disk	Disk Management Notion	May 19, 2021	Alan Jo	Alan Jo	May 19, 2021		# Anatomy of hard-disk - platters and double of heads of platters - since we can use upper and lower of platter - usually sector is 512 bytes - head need physical take off and landing - slow - Disk address = (driver, surface, track, sector) - Cylinder-based mapping - Seek time : The head is moved to the target track - latency : Rotational latency - Transfer time: Read/write the sector ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2Fcee3253a-c7b1-499f-accc-489628ef3d46%2Ftemp.jpg?table=block&id=ab97e574-2861-4904-90ea-f9a11fccce0d&cache=v2) ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F6578359c-9bba-4b35-80a4-ba9b4bd1ce17%2Ftemp.jpg?table=block&id=130aa835-3424-4abc-85c4-da18252e6633&cache=v2) ### Mapping algorithm or disk addressing <details><summary> Mapping from sector address to logical block address</summary> (surface j, track i, sector k) = k + s * (j + i*p) - Assuming p surfaces in the disk and s sectors per track. </details>	e4eca2717d6145a189df99f477adf720
Disk Allocation methods	Disk Management Notion	Nov 5, 2019	Alan Jo	Seong-lae Cho	May 19, 2021		### where should we allocate(save) in disk - contiguous allocation - has hole waste to cleanblock also has same sizesimple but - linked allocation (dynamic) - pointerif bad sector broken easily occur - indexed allocation - Has index blockhas pointer > do not trust memory - especially secondary memory > contiguous : 인접한	4a49727bed65460cb4db6426038b97b4
Disk caching	Disk Management Notion	Nov 5, 2019	Alan Jo	Seong-lae Cho	May 19, 2021		for slow disk performance Buffer cache (in memory) : cache file blocks in memory to capture locality in buffer cache or buffer cache - Idea: cache file blocks in memory to capture locality in buffer cache (or buffer cache) - this cache is system wide, shared by all processes - cache consistency issues can be occur Page cache - – Caches pages rather than disk blocks using virtual memory techniques. – Memory-mapped I/O uses a page cache. Unified Buffer cache -– A unified buffer cache uses the same page cache to cache both memory-mapped pages and ordinary file system I/O. – Avoids double caching and the possibility of inconsistencies between page caches and buffer caches.	f0d66b9112c24fbd9b6bb7f115da6af0
Disk Head Scheduling	Disk Management Notion	May 19, 2021	Alan Jo	Alan Jo	May 19, 2021		More than one request in the disk queue - read or write ### Algorithms - FCFS - fair but ineffcient - fifo - SSTF - starvation possible - select minimum seek time - usual algorithm - SCAN - elevator algorithm - possibility of double time - double direction - C-SCAN - one direction since disk is round - usual algorithm - provide uniform wait time - LOOK - one direction until services all requests → reverse - C-LOOK - variant of C-SCAN - only goes as far as the last request in each direction	e6e60cca761147059a1bf7717616c398
Disk layout	Disk Management Notion	Nov 5, 2019	Alan Jo	Seong-lae Cho	May 19, 2021		cpu can't directly control disk so kernel has in-memory sturcture ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2Fcf347120-7912-4dae-ae82-0b858e27757f%2Ftemp.jpg?table=block&id=cdd3a113-0151-4967-853c-5c59b1663de5&cache=v2) ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F8ba5baf0-04e8-425e-8fef-ec74cb81ac66%2FUntitled.png?table=block&id=79a04b2e-946f-4b24-93db-68bf497c9463&cache=v2) ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F7338d4d4-8caa-4323-bf34-c27267abf197%2FUntitled.png?table=block&id=74744e50-5351-46a0-b807-98e79499f478&cache=v2) ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F05994a8d-f0f0-4362-8359-d7984a89eab2%2FUntitled.png?table=block&id=eb9b0dd1-387a-4de2-ba9f-8f47759f634d&cache=v2)	5091a147a01a4ff2986c9d80e6aad326
Disk management	Disk Management Notion	May 19, 2021	Alan Jo	Alan Jo	May 19, 2021		[Disk partitioning](https://texonom.com/disk-partitioning-129115f572724f5cbc570c7517151187) [Disk bad sector](https://texonom.com/disk-bad-sector-5cc7aac350854f0885cfa60f62bee9ce) [Disk Swap Space](https://texonom.com/disk-swap-space-5bf2726b42c14bf1abeeb90018408163)	2db57b8f7837407088208b308e51bbec
Disk Performance	Disk Management Notion	May 19, 2021	Alan Jo	Alan Jo	May 19, 2021		### Performance - 1. cost, 2. speed & bandwidth, 3. reliability optical disk - permanent data storage → cloud in temporary ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2Fc5563700-295a-4a9d-8f1e-604d5a19d691%2FUntitled.png?table=block&id=8e30c91b-677f-420a-93ea-e7e39f363dd2&cache=v2) ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2Fda675d40-dc07-49e5-b1a5-5505aeed791b%2FUntitled.png?table=block&id=dd0f8f2c-c868-4f57-b9b1-0520b0e86759&cache=v2)	28e569d80cbe4c3b9e68857ac92550ea
RAID	Disk Management Notion	Nov 5, 2019	Alan Jo	Seong-lae Cho	Jun 7, 2021		## redundant array of inexpensive disks [Hardware RAID](https://texonom.com/hardware-raid-861083a0395649d18fa43e8b5dec9f51) ### RAID level - originally 5 level [RAID 0](https://texonom.com/raid-0-5b52502a6406464689c2bf557f5d3b47) [RAID 1](https://texonom.com/raid-1-b4b2a03a0193424e8fa40b4ee7488be7) [RAID 2](https://texonom.com/raid-2-1b24579f29674f05a13c00b5075f2739) [RAID 3 ](https://texonom.com/raid-3-4cc5f5d7b3c84145825a3f9bc4975ec9) [RAID 4](https://texonom.com/raid-4-f4db629dc21d4e58a8ad908a83be049f) [RAID 5](https://texonom.com/raid-5-afdf3a4add0a4374958c915fea1f83ef) Usual tech of server and NAS and data center ### Two design goals of various RAIDs 1. increased data reliabilitymirroring data 2. increased input/output performance trade-off between reliability and performance > even parity, odd parity for detect block failure > Striping : Disk striping - look at the diagrams which are like stripes	4b50ffdd255d430ca76f393e8d1ebe25
Tertiary storage devices	Disk Management Notion	Nov 5, 2019	Alan Jo	Seong-lae Cho	May 19, 2021		ex. usb, sd card, external disk [HSM](https://texonom.com/hsm-95b2b5fc37ca4982aed9c39ffe45ecb2)	def6411e9adf4220811b65706895e6e3
Disk bad sector	Disk management	null	null	null	null	null	[Disk bad sector remapping](https://texonom.com/disk-bad-sector-remapping-f18741071ab949c29a01c238403a0630)	5cc7aac350854f0885cfa60f62bee9ce
Disk partitioning	Disk management	null	null	null	null	null		129115f572724f5cbc570c7517151187
Disk Swap Space	Disk management	null	null	null	null	null	temporal image space of main memory ### Swap File approach ### Swap partition Approach - Swapping By File system read but slow because of using VFS but do not need partition - or do not need file system just image dump to this sector > rule of the thumb - twice of main memory	5bf2726b42c14bf1abeeb90018408163
Disk bad sector remapping	Disk bad sector	null	null	null	null	null		f18741071ab949c29a01c238403a0630
Hardware RAID	RAID	null	null	null	null	null	하드웨어 디바이스를 통해 RAID를 수행하는 경우 운영체제와 파일시스템은 RAID 메커니즘을 인식하지 못함 전용 하드웨어인 RAID card 자체에서 오류가 발생하는 경우 전체 디스크 array가 무용지물이 되는 경우도 > [네이버 클라우드 플랫폼 (NAVER Cloud Platform) : 네이버 블로그](https://blog.naver.com/n_cloudplatform/222384170516)	861083a0395649d18fa43e8b5dec9f51
RAID 0	RAID	null	null	null	null	null	![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F82335579-cbcb-400a-8260-2f4bcbb2ec06%2Ftemp.jpg?table=block&id=f24a948f-aad6-45fc-bd58-2d347fbc8e6c&cache=v2) - RAID 0 - striped disks	5b52502a6406464689c2bf557f5d3b47
RAID 1	RAID	null	null	null	null	null	![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F82335579-cbcb-400a-8260-2f4bcbb2ec06%2Ftemp.jpg?table=block&id=49941f40-28e1-47fd-9f36-f7f78bd839f9&cache=v2) - RAID 1 - disk mirroring	b4b2a03a0193424e8fa40b4ee7488be7
RAID 2	RAID	null	null	null	null	null		1b24579f29674f05a13c00b5075f2739
RAID 3	RAID	null	null	null	null	null	![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2Faafcb581-cf0c-4683-9ed0-c28bf5c9f412%2Ftemp.jpg?table=block&id=48020019-2481-4287-8a7e-3d635536caf4&cache=v2) - RAID 3 - Striped set with dedicated parityidea of redundant disk to increase reliability - many disk decrease reliabilitybut update parity block overhead and bottle neck to parity disk	4cc5f5d7b3c84145825a3f9bc4975ec9
RAID 4	RAID	null	null	null	null	null		f4db629dc21d4e58a8ad908a83be049f
RAID 5	RAID	null	null	null	null	null	![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2Ffeb998d5-2d36-4dc7-bf82-eb682687d18e%2Ftemp.jpg?table=block&id=a2fc4fbf-b4bb-419f-b0a9-2b231d62e11d&cache=v2) - RAID 5 - Striped set with distributed parity	afdf3a4add0a4374958c915fea1f83ef
HSM	Tertiary storage devices	null	null	null	null	null		95b2b5fc37ca4982aed9c39ffe45ecb2
Memory Address Changing	Memory Management Notion	Nov 5, 2019	Alan Jo	Seong-lae Cho	May 19, 2021		- Physical address space - Starting at address 0, going to address MAXsys - Logical address space - Starting at address 0, going to address MAXprog CPU(code) execute instructions by logical address > relative address is an example of logical address ### Address Generation ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2Fd35ce2bf-0ec6-44eb-a128-e63e6089a756%2FUntitled.png?table=block&id=2d6934fd-f96f-4c0d-a731-09f64ece4df1&cache=v2) until linking, code has relative address and after loading to memory, it has physical address → Hardware for Address Translation (For adequate performance) ← dynamically “relocated” - Base (also called Relocation) Register - used for validation in CPU (either limit R) - Limit Register - size of program ## Linking combining various pieces of code and data into a single file that can be loaded into memory <details><summary> Linker’s tasks</summary> 1. Symbol resolution : each global symbol in an object file is bound to a unique definition 2. Relocation : the physical memory address for each symbol is determined and where references to those objects are modified either </details> <details><summary> kind of linker</summary> 1. static linker(`linkage editor’) - Object modules combine while Compile time 2. dynamic linker - Useful with language libraries (no duplicated copies) - done in Load time or Run time </details> ### Shared library - Static libraries disadvantages - Potential for duplicating- ex. libc.a - automatically occur when executable - also occur by user code dlopen() - routines can be shared by multiple processes ex. [libc.so](http://libc.so)	53916e26496843d6bf5a26e6185ebe8d
Memory Management History	Memory Management Notion	May 19, 2021	Alan Jo	Alan Jo	May 19, 2021		# Backgrounds Multi programming by time sharing Need multiple processes in memory at once (number of I/O + CPU) So we need protected updating Memory management = The task that is carried out by the “OS and hardware(MMU)” to accommodate multiple processes and fully use memory resources to maximize performance Requirement - Each process should have a logically contiguous space → virtual memory, make user do not consider memory while make program ### Goal - protect, performance(fully use) and sharing	baac71af6c7c4281894e15349bde6777
Memory Management Method	Memory Management Notion	May 19, 2021	Alan Jo	Alan Jo	May 19, 2021		### Memory Management Methods \|Title\| \|:-:\| \|[Virtual Memory](https://texonom.com/virtual-memory-07bf10ba636b42e9818c009b17f893f0)\| \|[Paging](https://texonom.com/paging-0571138c29914318a5f6a717cc2b36c0)\| \|[Dynamic Partitioning](https://texonom.com/dynamic-partitioning-ec4c829bbfec46d28f441044264eb5a7)\| \|[Fixed partitioning](https://texonom.com/fixed-partitioning-068a5eb6e57b4146926dc02fc828b649)\| \|[Segmentation](https://texonom.com/segmentation-9927cad8b8bf4366abdc591e533e1060)\|	040223391bea4c148bcfee4f4064572f
Dynamic Partitioning	Memory Management Methods	Nov 5, 2019	Alan Jo	Seong-lae Cho	May 19, 2021		- make partition when process input come - Each process is allocated exactly as much memory as it requires - Has external fragment (hole is external with no internal) - Use compaction to shift processes to free memory by external (less use by i/o problem) > Compaction is possible only if relocation is dynamic, and is done at execution time > I/O problem - Do I/O only into Operating System buffers <details><summary> Possible algorithms of placement</summary> 1. Best-fit - The fragment as small as possible 2. First-fit 3. Next-fit: choose first hole from last placement 4. Worst-fit: allocate the largest hole </details> > Overlays - choose memory layout by program call graph to efficient memory distribute by function module code	ec4c829bbfec46d28f441044264eb5a7
Fixed partitioning	Memory Management Methods	Nov 5, 2019	Alan Jo	Seong-lae Cho	May 19, 2021		Partitions can be of equal or unequal sizes set by hardware - has internal fragmentation (Unequal-size partitions lessens these) - A program may be too large → design the program with overlays ### queue(s) fixed partitioning Placement algorithm with “unequal size” partitions use of multiple queues(minimize internal fragmentation) → some queue will empty → use of single queue (increase internal fragmentation) > internal means internal of partition	068a5eb6e57b4146926dc02fc828b649
Paging	Memory Management Methods	Nov 5, 2019	Alan Jo	Seong-lae Cho	May 19, 2021		### Physical address space of a process can be non-contiguous (Pages are contiguous) - Keep track of all free frames - internal fragment - Set up a page table to translate logical to physical addresses - A virtual (logical) address is a pair (p, o) - Physical memory address is a pair (f, o) - Not all pages are mapped to frames 내부 단편화 O, 외부 단편화 X 고정된 영역인 페이지(page)로 분할 ### Paging Notion \|Title\| \|:-:\| \|[Page Replacement Algorithm](https://texonom.com/page-replacement-algorithm-3721241bd6b84fd9b55960f82ecf1fbe)\| \|[Paging Table](https://texonom.com/paging-table-b368ef8543cf4628b837232f4fd7fe7c)\| \|[Associative Register](https://texonom.com/associative-register-538b0ad40e84419795da36b76959786f)\| \|[Paging Protection ](https://texonom.com/paging-protection-b0940a815ff94260bcae736355499b4b)\| \|[Hierarchical paging](https://texonom.com/hierarchical-paging-2b45d0a7145b4ae98735812f9f771786)\| \|[Segmentation Paging](https://texonom.com/segmentation-paging-dfb1e2f233f1401daf3f8dc74466a57e)\| \|[Buddy System](https://texonom.com/buddy-system-b83b6f21d398442a84ff9c4fcd2ce295)\| ### Paging Usage \|Title\| \|:-:\| \|[Paging Example](https://texonom.com/paging-example-35326cf37c694b23ab5bab18de683b9f)\| > [세그먼테이션(segmentation) VS 페이징(paging)](https://blog.naver.com/PostView.nhn?blogId=qbxlvnf11&logNo=221368476938&parentCategoryNo=&categoryNo=&viewDate=&isShowPopularPosts=false&from=postView)	0571138c29914318a5f6a717cc2b36c0
Segmentation	Memory Management Methods	Nov 5, 2019	Alan Jo	Seong-lae Cho	May 19, 2021		### more classical method before paging but used in virtual address image 가변적인 영역인 세그먼트(segment)로 내부 단편화 X, 외부 단편화 O - like paging - it has segment table also in each process Logical address consists of a two tuple - <segment-number, offset> Segment-table base register (STBR), Segment-table length register (STLR) exist - has segment number s(identify segment) and d index(of process entry)- Maps two-dimensional physical addresseseach entry has base and limit ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2Fa58476a4-a0c8-4077-aa61-4aa199746880%2FUntitled.png?table=block&id=2c19b285-1a31-4664-8470-b281f9517a36&cache=v2) Has also relocation, sharing, allocation issue for fragmentation ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F74859a9e-2904-49ab-80d3-450c8e24a75c%2FUntitled.png?table=block&id=4927863d-e58e-41d0-9615-2fa88e16f7fa&cache=v2) > [세그먼테이션(segmentation) VS 페이징(paging)](https://blog.naver.com/PostView.nhn?blogId=qbxlvnf11&logNo=221368476938&parentCategoryNo=&categoryNo=&viewDate=&isShowPopularPosts=false&from=postView)	9927cad8b8bf4366abdc591e533e1060
Virtual Memory	Memory Management Methods	Nov 5, 2019	Alan Jo	Seong-lae Cho	Jan 18, 2023		## Demand paging = virtual memory ### Motive : In any hardware, same code(address) should have same data (memory management by maintaining logical semantic) virtual address: logical address Concept : separating of user logical memory from physical memory > - Symbolic Address : access variable by name > - address binding : variable mapping to address and physical memory when compile Insight : program do not need all of their code and data at once, so we can separate (but of course have overhead) > - Implementation : demand paging effect : VM enables program to execute without requiring their entire address space to be resident in physical memory. isolate each process's memory, enable share files and address spaces. ### Virtual Memory Notion \|Title\| \|:-:\| \|[Page Table](https://texonom.com/page-table-ffbce6c10d9a4a8993bd0d809f4a4249)\| \|[Segmentation Fault](https://texonom.com/segmentation-fault-b1008afd3e3a4117a92c978ba3dc0c14)\| \|[Page Replacement](https://texonom.com/page-replacement-936ec615a0554950bc1ab233fb67434f)\| \|[page size](https://texonom.com/page-size-c50803e5e52b492c847de02404f0be36)\| \|[TLB reach](https://texonom.com/tlb-reach-afd1dceecb9542c086eab1d89b938824)\| \|[Swapping](https://texonom.com/swapping-98903f7a115e48808b17fc418468c8e3)\| \|[Page fault handler](https://texonom.com/page-fault-handler-b4d166a24a7740a9bc153f80701a03bc)\| \|[Virtual Memory Thrashing](https://texonom.com/virtual-memory-thrashing-f57fa2a2a76244278cca8ad66badd63f)\| \|[i/o interlock](https://texonom.com/io-interlock-458727109a584d01b1087036dd6922bc)\| \|[Pages Arguments](https://texonom.com/pages-arguments-287308a9ec104a0eb0c4e01b157c8831)\| \|[TLB (Translation Lookaside Buffer)](https://texonom.com/tlb-translation-lookaside-buffer-fd5ffc7e77c44f57a744915cc426c27e)\| ### Virtual Memory Notion Usages \|Title\| \|:-:\| \|[Linux Paging](https://texonom.com/linux-paging-dddeb1bb20e64c67bf7d38ae744e625c)\| \|[COW(copy-on-write)](https://texonom.com/cowcopy-on-write-b41107af21f04feb923f02eeaf677017)\| \|[Memory-Mapped Files](https://texonom.com/memory-mapped-files-09442c8a229e475083829ddf52eca697)\| > [8강. 메모리 관리](https://marlinbar.tistory.com/68?category=300896)	07bf10ba636b42e9818c009b17f893f0
Associative Register	Paging Notion	Nov 5, 2019	Alan Jo	Seong-lae Cho	May 19, 2021		parallel search - if associative register exist <details><summary> Address translation (A´, A´´) </summary> – If A´ is in associative register, get frame number out – Otherwise, get frame number from page table in memory </details> ### Effective Access time Associative lookup = `math: \epsilon `time unit - related to hardware performanc Assume memory cycle time is 1 microsecond Hit ratio (`math: \alpha`) - related to algorithm $$EAT = (1 + \epsilon)\alpha + (2 + \epsilon)(1 - \epsilon) \\ = 2 + \epsilon - \alpha$$	538b0ad40e84419795da36b76959786f
Buddy System	Paging Notion	Nov 5, 2019	Alan Jo	Seong-lae Cho	May 19, 2021		- used with virtual memory system either - combination of fixed and dynamic Compromise to overcome disadvantages of both fixed and dynamic partitioning schemes <details><summary> Alogorithm</summary> 1. Start with entire block 2. When request of size S is made > ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2Fa14d3c04-3016-4f0e-a409-5db1a488c95a%2FUntitled.png?table=block&id=1372aeb8-b364-4946-8e46-1cc0c6ac5fc5&cache=v2) 3. this process is repeated until the smallest block greater or equal to S is generated. 4. Two buddies are coalesced whenever both of them become de-allocated. > Given the address and size of a block, the address of buddy is automatically determined. The address of a block and its buddy differ in exactly one bit position > binary branch framework and buddy(share one parent) is has different bit of final address </details> ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F95bed60f-0f6b-4160-993d-7960c4d49888%2FUntitled.png?table=block&id=f3fd7300-747a-4204-b62f-8b947fe5c206&cache=v2) > modified form is used in UNIX SVR4 (and also in Linux) for kernel memory allocation (KMA) > Smallest size of block is page ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F4085edf3-c409-4719-b832-70c192b29e91%2FUntitled.png?table=block&id=5c4d8207-1c56-4bf2-9233-3a534cbe810d&cache=v2)	b83b6f21d398442a84ff9c4fcd2ce295
Hierarchical paging	Paging Notion	Nov 5, 2019	Alan Jo	Seong-lae Cho	May 19, 2021		– A simple technique is a two-level page table – Break up the logical address space into multiple page table Has page offset either but has multi page number in logical address ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F8138952b-d75a-443f-92a3-5dfbabd137b2%2FUntitled.png?table=block&id=5b07a593-5c04-4e1e-a521-806dc03a80fe&cache=v2) last page number indicate page of page table,other is root or outer page table	2b45d0a7145b4ae98735812f9f771786
Page Replacement Algorithm	Paging Notion	May 19, 2021	Alan Jo	Alan Jo	May 19, 2021		### Page Replacement Algorithms \|Title\| \|:-:\| \|[LRU Algorithm](https://texonom.com/lru-algorithm-1f181112165f4df8b5a5cdea967d65c1)\| \|[Clock Algorithm](https://texonom.com/clock-algorithm-80f852e82a0548f5856f09baf11f6cf1)\| \|[Second Chance Algorithm](https://texonom.com/second-chance-algorithm-afdfb27ecaac4854902ba38eec679e41)\| \|[FIFO Algorithm](https://texonom.com/fifo-algorithm-80eb13636a8f4edc8861c04033c9d637)\| \|[LFU Algorithm](https://texonom.com/lfu-algorithm-db665029bdb84a42a98099c471ec744b)\| > [우주를 위해 : 네이버 블로그](https://blog.naver.com/oidoman/90097323379) > [LRU 알고리즘 (Least Recently Used Algorithm)](https://j2wooooo.tistory.com/121)	3721241bd6b84fd9b55960f82ecf1fbe
Paging Protection	Paging Notion	Nov 5, 2019	Alan Jo	Seong-lae Cho	May 19, 2021		Memory protection bits - Implemented by associating protection bit with each page - e.g. read-only, read-write, in page table valid/invalid bit - changed by other page in same frame ### Sharing is easy - just indicate same frame and make it both valid	b0940a815ff94260bcae736355499b4b
Paging Table	Paging Notion	May 19, 2021	Alan Jo	Alan Jo	May 19, 2021		### Paging Tables \|Title\| \|:-:\| \|[Hashed page table](https://texonom.com/hashed-page-table-5503e136b5d94731911edb2a21ccd4c9)\| \|[Inverted page table](https://texonom.com/inverted-page-table-25d3365925cf4aea9b9eb834c1c6bbad)\| ### One table (mean one image) for each process: part of process’s state <details><summary> Contents</summary> 1. Flags: valid/invalid (also called resident) bit, dirty bit, reference (also called clock or used) bit used for clock page change algorithm 2. Page frame number </details> Page table is kept in main memory - Page-table base register (PTBR) points to the page table of each process - Page-table length register (PTLR) indicates size of the page table of each process > use tlb to lessen memory access overhead to find page page table size is pretty big → – Only need to map the portion of the address space actually being used (tiny fraction of entire address space) Approach Methods > – Multi-level(Hierarchical) paging, Hashed page table, Inverted page table	b368ef8543cf4628b837232f4fd7fe7c
Segmentation Paging	Paging Notion	Nov 5, 2019	Alan Jo	Seong-lae Cho	May 19, 2021		- Use segments to manage logically related units - Use pages to partition segments into fixed size chunks - first go to segment table (protection easy) > Need to allocate page table entries only for those pieces of the segments that have themselves been allocated. > ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2Fd370d686-0d0a-4b36-9304-1c52bb619bfa%2FUntitled.png?table=block&id=33d0e559-e0c3-4c22-9dcb-38bbc8eeb185&cache=v2) > ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F5c5ffdbb-d91c-4b65-b989-fde20a0389cf%2FUntitled.png?table=block&id=69399c6b-6efa-4b2e-a6bc-3665fc417f01&cache=v2) Logical (virtual) address → Linear address → Physical address (intel 32 architecture) ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2Fd6130a43-2117-4675-acf7-49442a5ed895%2FUntitled.png?table=block&id=8bd0a58e-360f-499a-b136-4a2bd00ce504&cache=v2) Segmentation: logical partition of memory - Segment Selector bits exist (segmentation register store segment selector bits usually) ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2Ff8645193-e5ef-4450-a3af-26da99b9b36c%2FUntitled.png?table=block&id=2b2edf19-d47f-4964-9766-a221553cf019&cache=v2) > Offset means relative address	dfb1e2f233f1401daf3f8dc74466a57e
Clock Algorithm	Page Replacement Algorithms	May 19, 2021	Alan Jo	Alan Jo	May 19, 2021		second-chance보다 더 효율적인 FIFO버전의 알고리즘 queue의 페이지를 이동시키지 않는다. clock 알고리즘은 순환형 리스트를 유지하고, 가장 오래된 페이지를 가리키는 hand(iterator)를 가지고 있다 페이지 폴트가 발생했고 빈 프레임이 없다면, hand가 가리키고 있는 페이지의 참조비트를 검사 만약 참조비트가 클리어 상태라면 hand가 가리키고 있는 페이지가 새로운 페이지로 사용 그렇지 않으면 참조비트를 클리어시킨다. 그 다음 hand는 증가하고, 페이지가 교체될 때까지 이 과정이 반복 > [우주를 위해 : 네이버 블로그](https://blog.naver.com/oidoman/90097323379)	80f852e82a0548f5856f09baf11f6cf1
FIFO Algorithm	Page Replacement Algorithms	May 19, 2021	Alan Jo	Alan Jo	May 19, 2021		FIFO : 페이지가 주기억장치에 적재된 시간을 기준으로 교체될 페이지를 선정하는 기법 단점 : 중요한 페이지가 오래 있었다는 이유만으로 교체되는 불합리. 가장 오래 있었던 페이지는 앞으로 계속 사용될 가능성이 있음. [Belady's Anomaly](https://texonom.com/beladys-anomaly-d5dfd0e9467f4aa1a7ba5d584c4c6d0c)	80eb13636a8f4edc8861c04033c9d637
LFU Algorithm	Page Replacement Algorithms	May 19, 2021	Alan Jo	Alan Jo	May 19, 2021		LFU : 가장 적은 횟수를 참조하는 페이지를 교체 단점 : 참조될 가능성이 많음에도 불구하고 횟수에 의한 방법이므로 최근에 사용된 프로그램을 교체시킬 가능성이 있고, 해당 횟수를 증가시키므로 오버헤드 발생	db665029bdb84a42a98099c471ec744b
LRU Algorithm	Page Replacement Algorithms	May 19, 2021	Alan Jo	Alan Jo	May 19, 2021		## Least Recently Used Algorithm 가장 오랫동안 참조되지 않은 페이지를 교체	1f181112165f4df8b5a5cdea967d65c1
Second Chance Algorithm	Page Replacement Algorithms	May 19, 2021	Alan Jo	Alan Jo	May 19, 2021		FIFO의 변형된 형태의 알고리즘 FIFO와 같이 제거될 페이지를 queue의 첫번째 요소로부터 찾는다 그러나 즉시 페이지를 제거하는 것이 아니라, 그 페이지의 참조비트가 세팅되어 있는지를 체크한다 세팅되어 있지 않으면 제거하는 것이고, 그렇지 않으면 참조비트를 클리어해서 queue의 맨 뒤로 이동시 > [우주를 위해 : 네이버 블로그](https://blog.naver.com/oidoman/90097323379)	afdfb27ecaac4854902ba38eec679e41
Belady's Anomaly	FIFO Algorithm	null	null	null	null	null	FIFO(First In First Out)에서, 원래 페이지 프레임의 개수를 늘리면 page fault발생이 감소 해야 하나, 오히려 늘어나는 경우가 발생 > [벨레이디의 모순 (Belady's Anomaly)](https://m.blog.naver.com/cookatrice/50037414550)	d5dfd0e9467f4aa1a7ba5d584c4c6d0c
Hashed page table	Paging Tables	Nov 5, 2019	Alan Jo	Seong-lae Cho	May 19, 2021		- The virtual page number is hashed into a page table - This page table contains a chain of elements hashing to the same location - need hash function and page table like hash table ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F054a272e-ba5a-447a-b1fa-9f432abbc97b%2FUntitled.png?table=block&id=cdb4b8cb-54a6-4019-81e5-e68ca1ad2580&cache=v2) Virtual page numbers are compared in this chain searching for a match. If a match is found, the corresponding physical frame is extracted (fast)	5503e136b5d94731911edb2a21ccd4c9
Inverted page table	Paging Tables	Nov 5, 2019	Alan Jo	Seong-lae Cho	May 19, 2021		### 역 페이지 테이블 One entry for each real page of memory (frame) - Only one page table in the system - Entry consists of the virtual address of the page stored in that real memory location, with information about the process that owns that page - by pid, p → get i by hash table for limit one either ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F5dfe95bd-9a3d-4589-9252-dd6a7c0c0213%2FUntitled.png?table=block&id=e6e67336-8bae-4204-b6aa-13a3029a7c15&cache=v2) > [역 페이지 테이블 (Inverted Page Table) > 도리의 디지털라이프](http://blog.skby.net/%EC%97%AD-%ED%8E%98%EC%9D%B4%EC%A7%80-%ED%85%8C%EC%9D%B4%EB%B8%94-inverted-page-table/)	25d3365925cf4aea9b9eb834c1c6bbad
Paging Example	Paging Usage	May 19, 2021	Alan Jo	Alan Jo	May 19, 2021		# Paging in Hardware (Intel x86) – physical addresses of page tables (CR3 register store) - two level in intel (+ page directory) ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2Fce48bc62-3c64-4981-9923-613303ea798a%2FUntitled.png?table=block&id=14d204a8-6dd8-4444-8190-33197645aa4b&cache=v2) Enable by setting PG bit in CR0 register 4kb so 20 upper bits to find page table ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2Ff3239cab-044c-4eae-ab4b-7782ae6ba8bf%2FUntitled.png?table=block&id=217f7801-2b3e-4b75-a709-0b6d37ce1c00&cache=v2) ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F11c86ab6-4414-47d7-a314-c6c3b4610e28%2FUntitled.png?table=block&id=79f5e569-5327-4fb8-9bac-bff87c58639c&cache=v2) ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F4bdc9f84-8bc1-4229-ac66-ebbc796c9ba8%2FUntitled.png?table=block&id=c4967232-f546-42b6-b243-c52905d4b111&cache=v2) > Linux support Four-level paging for future ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2Fb10eea4f-ecbb-44d7-8b3f-dacb0d1b89e7%2FUntitled.png?table=block&id=b3f394dc-0641-4f29-b741-7f20dfb188f7&cache=v2) ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F77e637c7-abbb-490c-86a4-5238dc928c7b%2FUntitled.png?table=block&id=03cdfda0-1c14-4877-8aeb-b01535c0e193&cache=v2) ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2Ff031c764-89ce-4d2e-a42f-037c38dc9cbe%2FUntitled.png?table=block&id=7097a2b7-c336-4177-aa5e-97a8c05f3632&cache=v2)	35326cf37c694b23ab5bab18de683b9f
i/o interlock	Virtual Memory Notion	Nov 5, 2019	Alan Jo	Seong-lae Cho	Jan 18, 2023		- 커널로 문제해결 - Lock pages in physical memory using lock bits.	458727109a584d01b1087036dd6922bc
Page fault handler	Virtual Memory Notion	Nov 5, 2019	Alan Jo	Seong-lae Cho	Jan 18, 2023		### It is kind of interrupt if required PTE (entry) has invalid bit in page table than cause an exception to throw so go to OS(kernel mode), bring in missing page into physical frame, update PTE and restart instruction. -Page fault service time: service the page-fault interrupt + Read in the page + Restart the process - Effective access time: memory access time * (1 - p) + disk access time * p (p = probability of page fault) degradation: effective access time / memory access time (low p than low degradation) > Locality: dependent on page replacement policy > page change occur by not only by PFH, since predict page usage by machine learning	b4d166a24a7740a9bc153f80701a03bc
Page Replacement	Virtual Memory Notion	Nov 5, 2019	Alan Jo	Seong-lae Cho	Jan 18, 2023		> Belady's Proof (Oracle solution: ideal): Evicting the page that will not be used for the longest period of time minimizes the number of page faults # Counting Algorithms - LFU(Least frequently used)but do not reflect time variablerecent used victim - MFU(Most Frequently Used)consider recent used # Timing algorithms - FIFOSimple: single pointer sufficesBelady's Anomaly: more frame do not improve fifo algorithm performance - LRU (Least recently used)by counters or stack ### but lru is complicate ### LRU Approximation (real use) - Sampled LRU (hardware support)reference bit - can be cleared when full - Second-Chance AlgorithmLRU clock(Secondd chance) - clock bitclock hand - PTE reference - often use modified bit(dirty bit) either > Evaluate algorithm usually by running it on a particular string of memory references (reference string) and computing the number of page faults on that string Evict picked victim page* Page fault service time – Service the page-fault interrupt + Read in the page + Restart the process ### Allocation of Frames each process need minimum number of pages - Must provide enough frames to hold all the different pages that any single instruction can reference - local replacement - can not steal other process's frame - Fixed(Equal) allocation - each algorithm to process - global replacement - can steal other process's frame - Priority allocation > Proportional allocation : related to process size set frame number > Priority allocation : set process priority # Page replacement Examples 1. windows NTdemand paging with clustering(gist)process has working-set min, working-set maximum and if exceed then automatic working-set trimming process perform to restore free memory 2. solaris 2 > ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2Ff66a2160-4cbc-4f7c-95f5-593519e3a76b%2FUntitled.png?table=block&id=21bfeffc-ef6d-43a1-b507-7f31bce81cbd&cache=v2)	936ec615a0554950bc1ab233fb67434f
page size	Virtual Memory Notion	Nov 5, 2019	Alan Jo	Seong-lae Cho	Jan 18, 2023		- Provide Multiple Page Sizeshardware page size various so we can choose by softwarelarge: small page table, small page faultsmall: less handling overhead, small internal fragment	c50803e5e52b492c847de02404f0be36
Page Table	Virtual Memory Notion	Nov 5, 2019	Alan Jo	Seong-lae Cho	Jan 18, 2023		Each Process has Page Table, Page table entry(PTE) has valid, access bit ### Page Table Arguments 1. V = {0 ,1 , ..., N-1}: virtual address space 2. P = {0, 1, ..., M-1}: physical address space 3. N > M: difference that not comparable 4. MAP: V → P U {null} address mapping function 5. null if in Disk (if valid is 0): page fault ### Page Table Parameters log upper = lower 6. P = Page size (bytes): typically KBs 7. N = Virtual Address limit 8. M = Physical address limit <details><summary> virtual address = virtual page number(VPN) bit + page offset</summary> VPN acts page index in page table </details> physical address = physical page number bit + page offset address translation change previous one only (32 bit to other) > compare start by page table base register page table need protection and cache do not need because of address translation	ffbce6c10d9a4a8993bd0d809f4a4249
Pages Arguments	Virtual Memory Notion	Nov 5, 2019	Alan Jo	Seong-lae Cho	Jan 18, 2023		Virtual page and Physical page both exists, Page Table connect VP to PP - P = total number of virtual pages - N = total number of Physical pages 1. PP can be placed in Disk (Storage) 2. Different VP can point out same PP (ex. library) > if disk read occur than DMA(direct memory access) transfer and then by I/O controller there will be interrupt processor	287308a9ec104a0eb0c4e01b157c8831
Segmentation Fault	Virtual Memory Notion	Jun 1, 2021	Alan Jo	Alan Jo	Jan 18, 2023	[Segmentation](https://texonom.com/segmentation-9927cad8b8bf4366abdc591e533e1060)	## Segfault OS에 메모리 액세스 위반을 알리기 위해 메모리 보호 기능이있는 하드웨어에서 발생하는 오류 customer handler를 설치하여 사용하지 않으면 Linux의 커널에서 동작하고 있는 process signal 처리 segfault는 일반적으로 프로세스를 종료하고 적절한 ulimit 설정을 갖춘 코어 덤프를 생성 [Segfault RSP](https://texonom.com/segfault-rsp-9fa0f64a2de94e26906b7689c0590fd4) [Segfault RIP](https://texonom.com/segfault-rip-364a0188edf74a3e9f535a5f733bc996) 오류 값은 페이지 오류 오류 코드 비트의 비트 마스크 from `arch/x86/mm/fault.c` > [What is Segfault Message?](https://hslee77.github.io/linux,cloud/2021/05/07/LinuxSegfault-Event.html)	b1008afd3e3a4117a92c978ba3dc0c14
Swapping	Virtual Memory Notion	Nov 5, 2019	Alan Jo	Seong-lae Cho	Jan 18, 2023		A process can be swapped temporarily out of memory to a backing store - and then brought back into memory for continued execution Backing store - Fast disk that must provide direct access to these memory images bring a page into memory only when it is needed (page-level swapping) - OS do movement of pages between memory/disks - evicted pages go to disk if page is dirty to a swap file	98903f7a115e48808b17fc418468c8e3
TLB reach	Virtual Memory Notion	Nov 5, 2019	Alan Jo	Seong-lae Cho	Jan 18, 2023		- TLB Reach = (RLB Size) * (Page Size)가변 페이지 사이즈 이용해서 커널한테는 큰 페이지	afd1dceecb9542c086eab1d89b938824
TLB (Translation Lookaside Buffer)	Virtual Memory Notion	Nov 5, 2019	Alan Jo	Seong-lae Cho	Jan 18, 2023		usually fully associative cache since page table is in memory ### bits: valid + dirty(since it is cache and if write back) + tag(process ID + virtual page number) + physical page number + page offset	fd5ffc7e77c44f57a744915cc426c27e
Virtual Memory Thrashing	Virtual Memory Notion	Nov 5, 2019	Alan Jo	Seong-lae Cho	Jan 18, 2023		<details><summary> Since lack of free frame(low RAM), process > physical memory then page fault rate will be very high</summary> ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F42e6be14-9a28-4c41-9344-709852b69ee1%2Ftemp.jpg?table=block&id=7f7afe8b-044d-458c-af60-912fb3450ba1&cache=v2) </details> → os use all time to swapping Solution Method two - working set, ppf > ### 1. Working set, locality set + others for continity = working set *Working Set: set of actively used pages `math: \Delta` : working-set window : fixed number of page references (ex. 10,000 instruction) > ex. WS(t1) = {1,2,3,4,5} ,한꺼번에 반납, 대여 $$\Sigma WSSi = total demand frames$$ OS monitor direct D, and if D > total number of frames in physical them swap process then how to direct?: like sampled lru use reference bit* and *interval time* increasing bit will make performance better 2^bit per interrupt ### 2. Page-fault Frequency Allocation Window's method directly measure and control page-fault rate establish acceptable page fault frequency - too high rate → increase number of frames - too low rate → decrease number of frames	f57fa2a2a76244278cca8ad66badd63f
Segfault RIP	Segmentation Fault	null	null	null	null	null	명령 포인터 레지스터 값	364a0188edf74a3e9f535a5f733bc996
Segfault RSP	Segmentation Fault	null	null	null	null	null	스택 포인터 레지스터 값	9fa0f64a2de94e26906b7689c0590fd4
COW(copy-on-write)	Virtual Memory Notion Usages	Nov 5, 2019	Alan Jo	Seong-lae Cho	Jan 18, 2023		like fork(do not need to copy code and data) we do resource sharing efficiently image copy on write time, parent and child processes to initially share the same If either process modifies a shared page, only then is the page copied. ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F024bcd14-9529-4600-bd90-9c52a217c55e%2FUntitled.png?table=block&id=10970c8f-03c3-4373-9c16-b2c1748a6a49&cache=v2)	b41107af21f04feb923f02eeaf677017
Linux Paging	Virtual Memory Notion Usages	Nov 5, 2019	Alan Jo	Seong-lae Cho	Jan 18, 2023		### four-level paging for future hardware which has many three level to mapping to hardware each process has its own Page Global Directory and its own set of page (upper, middle, null)tables ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F81e93678-3bfb-4c92-ad67-8febcbb65be7%2Ftemp.jpg?table=block&id=0c97af39-ebe2-4696-b377-c3c931e8f15f&cache=v2) ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F0d77a5e8-bea8-4ece-82bd-39352f2d73e0%2Ftemp.jpg?table=block&id=fad97902-3c20-4c54-a857-31c1235cc219&cache=v2) task struct data sturucture has mm pointer, mm is pointer which has image mm_struct data sturcture. that has mmap pointer. process's meta data sturcture vm araa struct pgd is page first table base address directory 1. read then page fault (segmentation fault) start end area 2. write code is fault so OS read commision 3. also fault because because r/w no physical in frame then replacement > ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2Fe3a42f71-e4d2-4057-97b8-80ace7bef2d1%2FUntitled.png?table=block&id=4dd69d90-4471-4c46-be35-7f144ff711d8&cache=v2) ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F56c9d36a-87cf-4f5b-b9ad-cbde6a4d621d%2Ftemp.jpg?table=block&id=67c9f0be-bd02-44ac-8eeb-48005baed1a4&cache=v2) KMA: dynamic memory allocation in head (3gb) by malloc. in kernel (1gb) memory allocationbecause kernel cannot use malloc - contiguous page frame allocator : alooc_pages() - noncontiguous page fram allocator : vmalloc() - memory object allocators : the slab alloctor- kmem_cache_alloc : congifuous allocation of kernel objects- kmalloc() : contiguous allocation of arbitrary length up to 128kb ## Page replacemnet in linux - two LRU listsinactive_list - pages not known to be importantactive_list - pages known to be referenced frequently - kernel swap daemon (a kernel thread) - more on page replacemnet mechanism > 64bit usually use 3 level paging	dddeb1bb20e64c67bf7d38ae744e625c
Memory-Mapped Files	Virtual Memory Notion Usages	Nov 5, 2019	Alan Jo	Seong-lae Cho	Jan 18, 2023		Memory-mapped file I/O allows file I/O to be treated as routine memory access by mapping a disk block to a page in memory - Uniform access for files and memory → less copying ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F26e99936-66b7-4614-ad26-1cc2b825c286%2FUntitled.png?table=block&id=02e0e36d-a529-43de-a147-3a10034222b8&cache=v2)	09442c8a229e475083829ddf52eca697
Garbage Collector	Memory Management Usage	Oct 29, 2020	Alan Jo	Alan Jo	May 19, 2021		### Garbage Collector Notion \|Title\| \|:-:\| \|[GC Compaction](https://texonom.com/gc-compaction-9179519ad2ec4ba48377ad941be50edf)\| \|[Generational GC](https://texonom.com/generational-gc-9e88cd3b020e417dbe4d0a231884d0c7)\| \|[Generational Theory](https://texonom.com/generational-theory-307b210f9661421580b4b22b331014c1)\| ### Garbage Collectors \|Title\| \|:-:\| \|[Static Garbage Collector](https://texonom.com/static-garbage-collector-eaac7f4e278741fe97fa212dd9037eba)\| \|[Dynamic Garbage Collector](https://texonom.com/dynamic-garbage-collector-7d88c0e0170b4c1c8a890b6e65932088)\|	8af9f92e3ca842ee8ddd638e25647cd7
Memory Leak	Memory Management Usage	Oct 29, 2020	Alan Jo	Alan Jo	May 19, 2021		1. Global Variable 2. Timer, Callback 3. Closer 4. Multi Reference > [자바스크립트와 V8 엔진의 메모리 관리 프로세스](https://medium.com/naver-place-dev/%EC%9E%90%EB%B0%94%EC%8A%A4%ED%81%AC%EB%A6%BD%ED%8A%B8%EC%99%80-v8-%EC%97%94%EC%A7%84%EC%9D%98-%EB%A9%94%EB%AA%A8%EB%A6%AC-%EA%B4%80%EB%A6%AC-%ED%94%84%EB%A1%9C%EC%84%B8%EC%8A%A4-f45091e696e1)	16c4f840b59e4d45a04770b03e582229
GC Compaction	Garbage Collector Notion	Apr 30, 2021	Alan Jo	Alan Jo	Apr 30, 2021	[Dynamic Garbage Collector](https://texonom.com/dynamic-garbage-collector-7d88c0e0170b4c1c8a890b6e65932088)	![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2Ff1cd8946-653d-4fa4-b9a8-d9d16a199728%2FUntitled.png?table=block&id=697cd2a4-4784-415e-b7dc-5f1465129ff1&cache=v2) > [Go 언어의 GC에 대해 - LINE ENGINEERING](https://engineering.linecorp.com/ko/blog/go-gc/)	9179519ad2ec4ba48377ad941be50edf
Generational GC	Garbage Collector Notion	Apr 30, 2021	Alan Jo	Alan Jo	Apr 30, 2021	[Generational Theory](https://texonom.com/generational-theory-307b210f9661421580b4b22b331014c1) [V8 GC](https://texonom.com/v8-gc-2630948f867a46069a56ccae5271f574) [JVM GC](https://texonom.com/jvm-gc-0023b0c7ffe644ed94a45f2a69fde1ea)	세대별 GC의 목적은 heap 내 객체를 수명(GC에서 살아 남은 횟수 등으로 나타냄)에 따라 분류하여 GC 효율을 향상시키는 것 - 객체 신규 할당 영역에선 GC를 자주 수행(Minor GC) - 해당 영역 GC에서 여러 번 살아 남은(가장 오래된) 객체는 승격(promote)시켜 GC 빈도가 낮은(Major GC) 영역으로 이동 Java8 HotSpot VM에서는 모든 컬렉터가 세대별 GC를 제공 [write barrier](https://texonom.com/write-barrier-73c82a9b955346429fee7fc504ab3172) > [Go 언어의 GC에 대해 - LINE ENGINEERING](https://engineering.linecorp.com/ko/blog/go-gc/)	9e88cd3b020e417dbe4d0a231884d0c7
Generational Theory	Garbage Collector Notion	Apr 30, 2021	Alan Jo	Alan Jo	Apr 30, 2021		새로 할당된 객체는 대부분 일찍 죽는다는 가설	307b210f9661421580b4b22b331014c1
write barrier	Generational GC	null	null	null	null	null	세대별 GC는 GC를 수행하지 않을 때도 애플리케이션에 오버헤드가 발생한다는 단점 root에서 신규 세대를 참조하는 경로만 조사한 뒤 그 중 접근 불가능한 상태인 것을 회수하면 그렇다고 오래된 객체를 포함하여 heap 전체를 검사해 버리면 세대별 GC를 하는 의미가 없다 그래서 애플리케이션에서 참조를 대입하거나 rewrite할 때, 오래된 세대에서 신규 세대를 참조하는 경우 해당 참조를 별도로 기록하는 처리가 추가 이와 같은 참조의 ‘mutate’와 함께 부수적으로 진행해야 하는 처리를 ‘write barrier’라고 ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F91b10137-4b01-4e06-8395-497fe3e92e6e%2FUntitled.png?table=block&id=65bf9e6d-3ddb-4cee-bf93-c23e71d43272&cache=v2) 세대별 GC를 사용하면 write barrier 오버헤드에 비해 얻을 수 있는 이점이 더 클 것으로 기대 > [Go 언어의 GC에 대해 - LINE ENGINEERING](https://engineering.linecorp.com/ko/blog/go-gc/)	73c82a9b955346429fee7fc504ab3172
Dynamic Garbage Collector	Garbage Collectors	Apr 30, 2021	Alan Jo	Alan Jo	Apr 30, 2021		GC 수행 시 alive 상태의 객체를 heap 끝으로 재배치해서 heap을 압축 - 단편화 회피 - bump allocation을 통해 고속 메모리 할당 구현 가능 > - 객체가 heap 끝에 위치하므로 메모리 신규 할당 시 끝에서 바로 increment ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2Ffd98b5ba-7490-479b-a026-7bcf994a6328%2FUntitled.png?table=block&id=1080d55c-9308-438e-bc75-e6e490419519&cache=v2) ### Dynamic Garbage Collectors \|Title\| \|:-:\| > [Go 언어의 GC에 대해 - LINE ENGINEERING](https://engineering.linecorp.com/ko/blog/go-gc/)	7d88c0e0170b4c1c8a890b6e65932088
Static Garbage Collector	Garbage Collectors	Apr 30, 2021	Alan Jo	Alan Jo	Apr 30, 2021		정적 유형의 GC는 GC별로 heap 내 객체를 재배치하지 않습니다 정적 유형 GC에서 메모리 할당과 해제를 반복하면 heap 단편화가 발생해 할당 성능이 악화되는 문제가 있다 단, 이는 메모리 할당자를 어떻게 구현하느냐에 따라 달라짐 ### Static Garbage Collectors \|Title\| \|:-:\| \|[Tracing garbage collection](https://texonom.com/tracing-garbage-collection-19e036fe4c964d12be25ea3a974b69fd)\| > [Go 언어의 GC에 대해 - LINE ENGINEERING](https://engineering.linecorp.com/ko/blog/go-gc/)	eaac7f4e278741fe97fa212dd9037eba
Tracing garbage collection	Static Garbage Collectors	Nov 9, 2020	Alan Jo	Alan Jo	Apr 30, 2021		[Stop the world GC](https://texonom.com/stop-the-world-gc-5b16f1ca6ac3450bbf63e3bf56397538) [Mark & Sweep GC](https://texonom.com/mark-sweep-gc-b7bb682c52bc4be98043c847381fe483) [Reference Counting](https://texonom.com/reference-counting-a780282e427c436f8e12ca7f95173926) [Strong and weak references](https://texonom.com/strong-and-weak-references-7c3454cf4f1843d991bdc68439ec0091) > [Tracing garbage collection](https://en.wikipedia.org/wiki/Tracing_garbage_collection#Stop-the-world_vs._incremental_vs._concurrent)	19e036fe4c964d12be25ea3a974b69fd
Mark & Sweep GC	Tracing garbage collection	null	null	null	null	null	### 루트에서 접근 가능한 메모리는 active, 나머지는 기비지 [CMS Collector](https://texonom.com/cms-collector-5893234d493d4a9a8e14b4a392d44827) active root 에 원치 않는 참조가 포함되어있다면, 필요 없는 메모리가 할당된 만큼 메모리 누수가 발생 2단계 가비지 컬렉션 알고리즘 1. Mark: GC 루트 목록을 생성합니다. 자바스크립트의 경우 `window`객체가 루트로 작동하는데, 루트란 코드에서 참조되는 전역 변수를 뜻합니다. 모든 루트들은 `acitve` 혹은 `alive`(가비지가 아님)로 표시되며, 이들의 자식들 또한 재귀적(recursively)으로 동일하게 처리됩니다. 즉, 루트에서 접근 가능하다면 가비지가 아니라고 판단됩니다. 2. Sweep: `active`로 표기되지 않은 메모리를 모두 가비지로 간주하여 반환합니다. > [자바스크립트와 V8 엔진의 메모리 관리 프로세스](https://medium.com/naver-place-dev/%EC%9E%90%EB%B0%94%EC%8A%A4%ED%81%AC%EB%A6%BD%ED%8A%B8%EC%99%80-v8-%EC%97%94%EC%A7%84%EC%9D%98-%EB%A9%94%EB%AA%A8%EB%A6%AC-%EA%B4%80%EB%A6%AC-%ED%94%84%EB%A1%9C%EC%84%B8%EC%8A%A4-f45091e696e1)	b7bb682c52bc4be98043c847381fe483
Reference Counting	Tracing garbage collection	null	null	null	null	null	객체를 참조하는 변수의 수를 추적하는 방법 레퍼런스가 하나 증가/복사 될 때마다 카운트가 1만큼 증가 레퍼런스 카운팅은 매우 심플하고 빠른 가비지 컬렉션 방식이지만, cyclic/circular reference 를 처리하지 못하는 등의 단점 스마트 포인터 cpp	a780282e427c436f8e12ca7f95173926
Stop the world GC	Tracing garbage collection	null	null	null	null	null		5b16f1ca6ac3450bbf63e3bf56397538
Strong and weak references	Tracing garbage collection	null	null	null	null	null		7c3454cf4f1843d991bdc68439ec0091
CMS Collector	Mark & Sweep GC	null	null	null	null	null	## Concurrent Mark & Sweep	5893234d493d4a9a8e14b4a392d44827
Interrupt	Processor Scheduling Notion	May 19, 2021	Alan Jo	Alan Jo	Mar 28, 2023	[reverie](https://github.com/facebookexperimental/reverie)	### interception ### Interrupt Notion \|Title\| \|:-:\| \|[Mode Switching](https://texonom.com/mode-switching-b7f85582f0474555a8c08da922b33610)\| \|[OS Mode](https://texonom.com/os-mode-b0375e3351064cbd9d220d1c9d44e0e8)\| \|[Syscall](https://texonom.com/syscall-74f2aa4e9d7141dbad05bae5c88a07e0)\| \|[Context Switching](https://texonom.com/context-switching-132451f1c9fe4324a1a2bd84e11e7696)\| \|[Preemptive Scheduling](https://texonom.com/preemptive-scheduling-c4b6e46752f14d69bb769e987ae1c21b)\| \|[Cooperative Scheduling](https://texonom.com/cooperative-scheduling-93d38df64c4843e68d71e5db4e0f5752)\| > [백엔드 개발자들이 알아야할 동시성 4 — 스케줄링](https://choi-geonu.medium.com/백엔드-개발자들이-알아야할-동시성-4-스케줄링-e684331afc77)	d26d34c54d5447a6888ad85089184fb3
Process	Processor Scheduling Notion	Nov 5, 2019	Alan Jo	Seong-lae Cho	May 19, 2023	[Memory Management](https://texonom.com/memory-management-dbf37c509eaf46deb2df8ccd086e386c)	### process - program in execution ### Process Notion \|Title\| \|:-:\| \|[Process Sync](https://texonom.com/process-sync-45211a317f4f442cbab6b374fdaed630)\| \|[Process switching ](https://texonom.com/process-switching-7f2486a6184241fba896860610a4d5f1)\| \|[Process Memory](https://texonom.com/process-memory-53285d27582d49058052939d739e8623)\| \|[Process Function Call](https://texonom.com/process-function-call-ab41c2af3a5743599fca591ce08b535f)\| \|[Process State](https://texonom.com/process-state-bbe0f9abe7b240c7a1d2a2d1c3478afd)\| \|[Process control block](https://texonom.com/process-control-block-f62e2db348e442c580ea6d2084973973)\| \|[Process descriptor](https://texonom.com/process-descriptor-81614898a94d48058419b80e9e971fae)\| \|[Process image ](https://texonom.com/process-image-c6384d00d1af41a6bcaa1b238165be17)\| \|[Process context](https://texonom.com/process-context-e76d70c66cd840b4a247de3310e85416)\| \|[Process Exit](https://texonom.com/process-exit-2d82b561fba64541bde82c478297a22c)\| \|[Critical Section](https://texonom.com/critical-section-8469b236528b41efb4088ea3a69a4cc4)\| \|[Semaphore](https://texonom.com/semaphore-46b11416972249d2a40e6c3e5060304a)\| \|[Monitor Sync](https://texonom.com/monitor-sync-54b03a3282964bdb88207ec60dfeb1b2)\| \|[IPC](https://texonom.com/ipc-98010a1077fd45968ce78ae98505b2e5)\| \|[massage passing](https://texonom.com/massage-passing-ab22575af94b441caad37d2e1697cd59)\| \|[Mutex](https://texonom.com/mutex-cbfff707d2fe484ea20344f1104b7141)\| > memory, device, files, process has each table > process image is get by pid > [서버에 걸리는 부하, 추측하지 말고 계측하자](https://injae-kim.github.io/dev/2020/07/09/how-to-check-single-server-load-average.html)	ffd68f9e2b2648759b96047d48d17778
Processor Scheduler	Processor Scheduling Notion	May 19, 2021	Alan Jo	Alan Jo	Feb 21, 2023		### Processor Schedulers \|Title\| \|:-:\| \|[Long-term scheduler](https://texonom.com/long-term-scheduler-5f0783472fa94c478b10b7a1bfbf0192)\| \|[Short-term scheduler](https://texonom.com/short-term-scheduler-f95f24cbd78d490c8a216dcabb04e4cb)\| \|[Medium-term scheduler](https://texonom.com/medium-term-scheduler-15b70c90c2f641e696fcc6c90cbd04d9)\| ### Processor Scheduler Notion \|Title\| \|:-:\| \|[Processor Dispatcher](https://texonom.com/processor-dispatcher-dac6ca05f13c4ddc82906bd73f6080df)\| ### Processor Queues \|Title\| \|:-:\| \|[Job Queue ](https://texonom.com/job-queue-92bccac0db114b5c83a0f3ac0d9de17e)\| \|[Ready Queue](https://texonom.com/ready-queue-c8546adf7d604a73a15b292808bef5b7)\| \|[Device Queue](https://texonom.com/device-queue-453d40088668423c96d51003b27f9a2e)\| - selection function (which) in ready Q - decision mode (when) - end time > - non preem > - preem > [[O/S] CPU Scheduler와 Dispatcher](https://velog.io/@ss-won/OS-CPU-Scheduler%EC%99%80-Dispatcher)	8c124741c1e84d26b0b0a8ad599337f4
Thread	Processor Scheduling Notion	Nov 5, 2019	Alan Jo	Seong-lae Cho	Feb 21, 2023		### Thread Notion \|Title\| \|:-:\| \|[Thread Safe](https://texonom.com/thread-safe-a6b8ab6ec6e146deafa86b428e5ecfbe)\| \|[Threads model](https://texonom.com/threads-model-800491618af24626b02f701d43578b66)\| \|[kernel level thead](https://texonom.com/kernel-level-thead-498f95b4e748492b8e2ed535bf987b0b)\| \|[Lock convoy](https://texonom.com/lock-convoy-a904af475dcb4bb8a0d99ea8e095de39)\| \|[Two-Step Dance](https://texonom.com/two-step-dance-b328c6d4bef8483a97e076bbc91b0be8)\| \|[Priority Inversion](https://texonom.com/priority-inversion-beb6115d6154481db836a5677f877581)\| \|[Threads Scheduling](https://texonom.com/threads-scheduling-e1315fdaa06740c6ab54f097edeba1d7)\| \|[Deadlock](https://texonom.com/deadlock-8c118f296cc1477aac5c59c75cec8235)\| \|[CAS](https://texonom.com/cas-b3541e0665b74cf7aa4f91a7b202fde2)\| ### Thread Types \|Title\| \|:-:\| \|[user level thread](https://texonom.com/user-level-thread-6f68e5d2dff240359b9b494534598b29)\| > [좀 생소한 스레드 관련 문제](https://m.blog.naver.com/PostView.nhn?blogId=jjoommnn&logNo=130036504092&proxyReferer=https:%2F%2Fwww.google.com%2F)	d704a229d3b94619963c61bb742b822b
**Context Switching**	Interrupt Notion	Mar 28, 2023	Alan Jo	Alan Jo	Mar 28, 2023		### **Context Switching Notion** \|Title\| \|:-:\|	132451f1c9fe4324a1a2bd84e11e7696
**Cooperative Scheduling**	Interrupt Notion	Mar 28, 2023	Alan Jo	Alan Jo	Mar 28, 2023		협조적 스케줄링	93d38df64c4843e68d71e5db4e0f5752
Mode Switching	Interrupt Notion	May 19, 2021	Alan Jo	Alan Jo	Feb 21, 2023			b7f85582f0474555a8c08da922b33610
OS Mode	Interrupt Notion	Nov 5, 2019	Alan Jo	Seong-lae Cho	Feb 21, 2023		# hardware protection 1. dual mode - user mode / kernel mode - need hardware support 2. hardware protect protected processor register in intel CPL(current), RPL(request), DPL(descriptor) > supervisor = system = monitor = kernel > before mode > intel real mode, protected mode # Mode change ### 1. hardware(external) interrupt - i/O interrupt - device interrupt ← timer intterupt ### 2. software(internal) interrupt - exception → exception handler ### 3. system call # system call → system call handler (by system call interface - inplement of wapper funct in libc.a) ### Method > int 0x80(vector) or sysenter > 1. same reg to KM stack > 2. invoke system call service routine > 3. exit handler > syscall_exit or sysexit > <details><summary> > > example</summary> > 1. fork () > 2. fork{} in libc.a > 3. movl 2, %eax - 2 means fork, eax is register > 4. int 0x80(system call) > 5. go to IDT(interrupt desciptor table) > 6. go to kernel mode function > 7. save all in stack > 8. check eax to function > 9. find sys call table that (2, fork) > 10. do ref_from_sys_call in entry.S > > </details> > > ### kind > > [System call](https://texonom.com/system-call-f36194e943f3498593520460b36c55aa) > - process management system call - file management system call - file system management system call > &system_call - handler address # Bridge there is breidge like north and south to control traffic > no bridge in embedded # VM - hosted - up on host OS - stand alone - up on hardware [dual booting](https://texonom.com/dual-booting-24656ee018014e6388b841b9f579c30c)	b0375e3351064cbd9d220d1c9d44e0e8
Preemptive Scheduling	Interrupt Notion	Mar 28, 2023	Alan Jo	Alan Jo	Mar 28, 2023		선점형 스케쥴링	c4b6e46752f14d69bb769e987ae1c21b
Syscall	Interrupt Notion	Feb 21, 2023	Alan Jo	Alan Jo	Feb 21, 2023			74f2aa4e9d7141dbad05bae5c88a07e0

**Claude Shannon**

Computer Scientists

May 9, 2023

Alan Jo

Jun 25, 2023

[Alan Turing](https://texonom.com/alan-turing-df61b5084e544ac58577d289983e0d28)

구글 학술검색 기준으로 세상에서 4번째로 많이 인용된 논문인 <정보통신의 수학적 이론>이라는 논문을 발표 [Entropy](https://texonom.com/entropy-35fdf9bb7fdd42edbd4277147ca0ec38) 1950~1986 연 평균 복리수익율이 28% 로 세계최고수준 [N gram Language Model](https://texonom.com/n-gram-language-model-e600ddf01e2645e1b1d76604454bc3c9) [One Time Pad](https://texonom.com/one-time-pad-0fd4b702a7854dca8263035014fa3914)

bc27d0c33de4438cb326957a169ed36a

Edsger Dijkstra

Computer Scientists

Aug 9, 2023

Alan Jo

Aug 9, 2023

> [Edsger Dijkstra's One-Day Workweek - Cal Newport](https://calnewport.com/edsger-dijkstras-one-day-workweek)

569bf90716e64b468e888f51099c8beb

Grace Hopper

Computer Scientists

Sep 7, 2020

Alan Jo

Jun 25, 2023

최초로 컴파일러를 만들고 처음으로 디버깅(?)을 한 그레이스 호퍼도 여성 1947년 호퍼는 마크2에서 프로그래밍 작업을 하고 있었는데, 그 날 따라 천공카드 입력이 잘 안되었다. 결국, 릴레이 안에 죽어있는 나방 때문에 컴퓨터가 제대로 동작하지 않는다는 사실을 알게 된 그리고 그 나방을 작업 노트에 붙이면서 최초의 컴퓨터 버그라고 메모를 남긴다.

1b108f44e96a4899b18c1a6821869f48

Margaret ***Hamilton***

Computer Scientists

Sep 7, 2020

Alan Jo

Jun 25, 2023

### software engineering 이라는 용어를 처음 사용했다 대학에서 수학을 공부한 그녀는 남편의 박사 학위 공부를 지원하기 위해 MIT에서 프로그래머로 일을 시작했다. 당시 그녀는 제대로된 프로그래밍 교육도 받지 못하고 바로 날씨를 예측하는 프로그램 개발에 참여 그 후, 시스템 프로그래밍 전문가가 되어 아폴로11 프로젝트에 참가하게 된다. 당시 아폴로11 프로젝트에는 SW개발을 위한 예산과 일정이 포함되어 있지 않았고, 요구 사항에도 소프트웨어 대한 언급은 없었다 > [F/OSS Story - 페이지 5 - 만화로 나누는 자유/오픈소스 소프트웨어 이야기](https://joonecomics.wordpress.com/category/foss-story/page/5/)

ea62b054dc0a47d291b80dc976b5f8d8

Shumuel Winograd

Computer Scientists

Jun 25, 2023

Alan Jo

Jun 25, 2023

5f38451bbad04b6da44260f93ca4eb43

Von Neumann

Computer Scientists

Apr 15, 2023

Alan Jo

Jun 25, 2023

ef79cb1b1e9045d28caced75663e14f7

Alan Turing History

Alan Turing Notion

May 9, 2023

Alan Jo

May 9, 2023

케임브릿지 대학교 입학 프린스턴 양자역학 논문에 영감을 받음 그 논문이 존 폰 노이만이 쓴 것 힐베르트가 기계가 만능이다라는 주장을 했고 세계수학학계가 암울했는데 학교에서 기계가 만능이 아니다라는 의미를 가지는 [불완전성 정리](https://texonom.com/e158a278de5b4a7fa97b509cbd2dc0bd) 를 읽고 연산을 수행하는 자동화된 기계라는 분야에 흥미를 가지고 공부하게 된다 괴델의 불완전성 원리의 간단함에 새로 자신만의 방식으로 증명 그 과정에서 universial machine이라는 단어 나옴 이런 방식으로 람다대수를 만들어서 불완전성 원리 증명한 알론소 처치를 따라 프린스턴으로 떠남 만들어냄 22살에 수학박사되고 논문 엄청 쓴 존 폰 노이만과 교류 맨하탄 프로젝트에 존 폰 노이만이 참여 앨런 튜링은 암호 얀그서 에드박 참여 논문 전쟁 끝나고 냄 근데 중단됨 못알아봐서 이후에 몇년뒤 구현되었는데 당시 세계젤빠른 컴퓨터 폰노이만이 edvac 보고서 미국서 썻는데 튜링머신 아이디어 참고했는지 확실하지 않다 이후 영국 현실에 막혀 맨체스터 대학에 가서 인공지능 공부 수학과 생물학 공부하며 논문 쓴다 과연 기계는 생각을 할수 있을까라는 질문으로 논문 시작 철학적 논문can machine think? Imitation Game 여기서 나오는 말 , 강화학습 개념 소개 모든 지성의 영역에서 인간과 기계의 경쟁을 예측 이후 자연에서 발생하는 패턴이 생기는 과정을 수학적으로 서술 호르몬 화학치료중 사과먹고 자살 폰 이만 구조 이론적 창시자 AI의 선두자이자 튜링 테스트의 어원 애니그마 개발자 어린시절 튜링은 친구를 되살리고 싶은 마음에 양자역학과 수학에 심취했다 엘런 튜링의 기계에 대한 증명은 괴델의 불확정성 원리에 기반한다 2차대전 에니그마 독일 이 uboat (undersea)에서 사용한 암호화기술 - 1943 - 앨런 튜링 에니악도 수소폭탄 시뮬레이션 위해 만들어진 것 - 1946 - 미국 폰 노이만이 edvac 보고서로 폰 노이만 구조 정립 (1936 앨런튜링이 제안한 논문과 거의 비슷 - 하지만 튜링 언급 없다) 엘란 튜링이 보니 불완전해보이는 보고서 → 영국에서 만들어주세요 해서 더 대단한 Proposed Electronic Calculator라는 논문을 낸다 Automatic Computing Engine 으로 이름짓는다 edvac을 넘어 서브루틴 가능하게 구조 설계했지만 너무 복잡하다고 영국에서 불가능한거 아니냐고 못알아먹음 ace프로젝트 중단됨 3년뒤 pilot ace라는 계속기로 제작되서 세계에서 가장 빠른 컴퓨터로 기록된다 그 후 궤델 소개해준 맥스뉴먼 따라 맨체스터 대학 가서 연구 모두 초월한 인공지능을 만들어야겠다 해서 수학적 생물학 공부 시작 1950년에 - computing machinery and intelligence를 만든다 The imitation game이라는 유명한 단어가 나온다 → turing test 라고 불리며 매년 세계에서 열리고 있다 2014년에 유진 구스트만이 최초로 튜링 테스트를 통과한다 이 논문 마지막으로 한문장으로 정리 - 철학적 논문 이후에 직접 turochamp를 개발하게 되지만 진짜 실행에는 실패 - 알고리즘 복잡 2012년도가 되서야 만든다 - 결과는 좋지 못했지만 morphogenesis라는 개념 자연에서 발생하는 패턴이 생성되는 과정을 수학적으로 설명하는 논문도 내게 된다 - 생물학 논문 아놀드 머레이라는 남자와 연인관계했지만 동성애 영국 불법이라서 구금된다 2013년에 영국은 동성애결혼이 합법화된다 호르몬치료를 받으며 슬픔을 못이겨 청산가리를 만들어 사과에 넣고 자살한다 이후 영국은 에니그마를 다른 국가에 보안 확실하다고 팔고 자기들만 해독해서 그 정보 사용한다 영국은 이를 감추기 위해, 동성애자 앨런 튜링을 감춘다 the ultra secret에서 튜링의 업적이 밝혀진다 영국과 애니그마 2009년에 앨런튜링 사면운동이 일어난다 - 2010년에 스티븐 호킹이 가세하고 10대 인물에 엘런 튜링이 들어가고 엘리자베스 2세 여왕의 특별사면권으로 공식 사면된다 컴퓨터계의 노벨상인 튜링상이 생긴다 > [앨런 튜링 - 현대 컴퓨터 과학의 아버지 (feat. 세계 2차 대전) [상]](https://www.youtube.com/watch?v=QasRpCP8qsI) > [2. 앨런 튜링과 폰노이만](https://joone.net/2016/11/15/%ec%95%a8%eb%9f%b0-%ed%8a%9c%eb%a7%81%ea%b3%bc-%ed%8f%b0%eb%85%b8%ec%9d%b4%eb%a7%8c/)

b394066bcd2a49fb9e048836bbf1a212

Automatic Computing Engine

Alan Turing Notion

Mar 2, 2021

Alan Jo

May 9, 2023

처음으로 메모리 내장 이용 cpu memory 입력장치 출력장치 설계

a61d7bc038fe4e58a80989d1c3e6facf

OS Function

OS Notion

May 19, 2021

Alan Jo

Jan 18, 2023

### OS Functions |Title| |:-:| |[Processor Scheduling](https://texonom.com/processor-scheduling-924d2fa78737494d82e74595de81219f)| |[Memory Management](https://texonom.com/memory-management-dbf37c509eaf46deb2df8ccd086e386c)| |[Disk Management](https://texonom.com/disk-management-f82047a4bc5a4bfb9960448cd13ac476)|

cc6a2b8fffb94340b51bc3f4a9c38862

OS History

OS Notion

Nov 5, 2019

Alan Jo

Seong-lae Cho

Jun 21, 2022

초기 컴퓨터들은 [Punch Card](https://texonom.com/punch-card-dcb71893cc074afb9ec6865b43eaaa61) 를 입,출력에 사용 물리적으로 컴퓨터 메모리에 로드해서 실행시키고 실행이 완료되면 메모리 덤프를 해서 출력은 얻음 [TTY](https://texonom.com/tty-d3c35ff3460343359b453c88a3bb896f) [Punch Card](https://texonom.com/punch-card-dcb71893cc074afb9ec6865b43eaaa61) ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2Fc55655c1-5711-44cc-be47-ee9da8f95035%2FUntitled.png?table=block&id=2248fd3f-a36c-4f4b-a903-2b3d3c4d3548&cache=v2) 이 과정을 사람이 직접 반복하다가 프로그램에 의해 자동화 하기 시작한것이 Operating System 의 시작 > unix: bsd (now free bsd), system V: bsd: for 학문: system V: ex. 솔라리스 >> init in bsd: getty >> init in system V: inittab >> those are psd 1 > > UNIX → GNU → Linux > → POSIX → BSD → MACOS > ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F8666dfa6-0c1f-4d3f-be0c-245ae45e6ee4%2FScreenshot_20200509-172659_Chrome.jpg?table=block&id=c42adbe0-47e2-4bd2-8ae4-33e1627aba09&cache=v2) > [Introduction](https://mug896.github.io/bash-shell/tty.html)

0f04c35c56144da9a9646d935a4a69fd

OS Structure

OS Notion

May 19, 2021

Alan Jo

Apr 3, 2023

[Software Architecture](https://texonom.com/software-architecture-5d3c80f6ffe44d60b10c870903c2b703)

a66cf2a9ea8f4841b64ca65fbec8b743

UNIX

OS Notion

Mar 18, 2021

Alan Jo

Apr 25, 2023

## **UNiplexed Information Computing System** ### Unix Notion |Title| |:-:| |[POSIX](https://texonom.com/posix-6fd6ba0425f145fca4dea48a6dee1ce7)| |[Unix Philosophy](https://texonom.com/unix-philosophy-8f96f33ed93b4ea5a533303869abfa92)| |[UNIX Structure](https://texonom.com/unix-structure-f892f7c631854084a7a93cf8db918cc3)| |[Unix-like](https://texonom.com/unix-like-578ba7dbb07b44b3a2d3895ca103c5ec)| |[UNIX History](https://texonom.com/unix-history-34c9169587714b508aad541945d7571d)|

233213b96425492f925a8277650c5b28

Disk Management

OS Functions

Nov 5, 2019

Alan Jo

Seong-lae Cho

Jan 18, 2023

[File System](https://texonom.com/file-system-7a340052fcaf48169a5c6da2610aa5cf)

### Need - Ram and disk price difference is still 10^2 - Memory is volatile - Disks as swap space for virtual memory system ### Disk Management Notion |Title| |:-:| |[Disk layout](https://texonom.com/disk-layout-5091a147a01a4ff2986c9d80e6aad326)| |[Disk Allocation methods](https://texonom.com/disk-allocation-methods-4a49727bed65460cb4db6426038b97b4)| |[Disk caching](https://texonom.com/disk-caching-f0d66b9112c24fbd9b6bb7f115da6af0)| |[Anatomy of hard-disk](https://texonom.com/anatomy-of-hard-disk-e4eca2717d6145a189df99f477adf720)| |[Disk Head Scheduling](https://texonom.com/disk-head-scheduling-e6e60cca761147059a1bf7717616c398)| |[Disk management](https://texonom.com/disk-management-2db57b8f7837407088208b308e51bbec)| |[RAID](https://texonom.com/raid-4b50ffdd255d430ca76f393e8d1ebe25)| |[Tertiary storage devices](https://texonom.com/tertiary-storage-devices-def6411e9adf4220811b65706895e6e3)| |[Disk Performance](https://texonom.com/disk-performance-28e569d80cbe4c3b9e68857ac92550ea)| > primary - secondary - tertiary

f82047a4bc5a4bfb9960448cd13ac476

Memory Management

OS Functions

Nov 5, 2019

Alan Jo

Seong-lae Cho

Jan 18, 2023

### Memory Management Notion |Title| |:-:| |[Memory Management History](https://texonom.com/memory-management-history-baac71af6c7c4281894e15349bde6777)| |[Memory Management Method](https://texonom.com/memory-management-method-040223391bea4c148bcfee4f4064572f)| |[Memory Address Changing](https://texonom.com/memory-address-changing-53916e26496843d6bf5a26e6185ebe8d)| ### Memory Management Usage |Title| |:-:| |[Garbage Collector](https://texonom.com/garbage-collector-8af9f92e3ca842ee8ddd638e25647cd7)| |[Memory Leak](https://texonom.com/memory-leak-16c4f840b59e4d45a04770b03e582229)| ### reference bit > [Untitled](https://asfirstalways.tistory.com/140) ### associative register > [연관기억장치, 복수 모듈 기억장치](https://devkim93.tistory.com/entry/%EC%97%B0%EA%B4%80%EA%B8%B0%EC%96%B5%EC%9E%A5%EC%B9%98-%EB%B3%B5%EC%88%98-%EB%AA%A8%EB%93%88-%EA%B8%B0%EC%96%B5%EC%9E%A5%EC%B9%98)

dbf37c509eaf46deb2df8ccd086e386c

Processor Scheduling

OS Functions

Nov 5, 2019

Alan Jo

Seong-lae Cho

Aug 10, 2023

### Scheduling is small issue in OS but hard to optimize ### Processor Scheduling Notion |Title| |:-:| |[Processor Scheduler](https://texonom.com/processor-scheduler-8c124741c1e84d26b0b0a8ad599337f4)| |[Process](https://texonom.com/process-ffd68f9e2b2648759b96047d48d17778)| |[Interrupt](https://texonom.com/interrupt-d26d34c54d5447a6888ad85089184fb3)| |[Thread](https://texonom.com/thread-d704a229d3b94619963c61bb742b822b)| ### Processor Scheduling Usages |Title| |:-:| |[Processor Scheduling Algorithm](https://texonom.com/processor-scheduling-algorithm-305e26d3aca84cdd9e5c980e91642a8f)| |[Processor Scheduling Method](https://texonom.com/processor-scheduling-method-c5c243a128b74b2183508330aec1090c)| |[Processor Scheduling Evaluation](https://texonom.com/processor-scheduling-evaluation-b897d00ffba948e6922c062d95693455)| > [[OS] 5-2. CPU Scheduling II](https://asfirstalways.tistory.com/124)

924d2fa78737494d82e74595de81219f

Anatomy of hard-disk

Disk Management Notion

May 19, 2021

Alan Jo

May 19, 2021

# Anatomy of hard-disk - **platters **and **double of heads of platters** - since we can use upper and lower of platter - usually **sector **is 512 bytes - head need physical take off and landing - slow - **Disk address = (driver, surface, track, sector)** - Cylinder-based mapping - Seek time : The head is moved to the target track - latency : Rotational latency - Transfer time: Read/write the sector ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2Fcee3253a-c7b1-499f-accc-489628ef3d46%2Ftemp.jpg?table=block&id=ab97e574-2861-4904-90ea-f9a11fccce0d&cache=v2) ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F6578359c-9bba-4b35-80a4-ba9b4bd1ce17%2Ftemp.jpg?table=block&id=130aa835-3424-4abc-85c4-da18252e6633&cache=v2) ### Mapping algorithm or disk addressing <details><summary> Mapping from sector address to logical block address</summary> (surface j, track i, sector k) = k + s * (j + i*p) - Assuming p surfaces in the disk and s sectors per track. </details>

e4eca2717d6145a189df99f477adf720

Disk Allocation methods

Disk Management Notion

Nov 5, 2019

Alan Jo

Seong-lae Cho

May 19, 2021

### where should we allocate(save) in disk - contiguous allocation - has hole waste to cleanblock also has same sizesimple but - linked allocation (dynamic) - pointerif bad sector broken easily occur - indexed allocation - Has index blockhas pointer > do not trust memory - especially secondary memory > contiguous : 인접한

4a49727bed65460cb4db6426038b97b4

Disk caching

Disk Management Notion

Nov 5, 2019

Alan Jo

Seong-lae Cho

May 19, 2021

for slow disk performance **Buffer cache (in memory)** : cache file blocks in memory to capture locality in buffer cache or buffer cache - Idea: cache file blocks in memory to capture locality in buffer cache (or buffer cache) - this cache is system wide, shared by all processes - cache consistency issues can be occur **Page cache** - – Caches pages rather than disk blocks using virtual memory techniques. – Memory-mapped I/O uses a page cache. **Unified Buffer cache** -– A unified buffer cache uses the same page cache to cache both memory-mapped pages and ordinary file system I/O. – Avoids double caching and the possibility of inconsistencies between page caches and buffer caches.

f0d66b9112c24fbd9b6bb7f115da6af0

Disk Head Scheduling

Disk Management Notion

May 19, 2021

Alan Jo

May 19, 2021

More than one request in the disk queue - read or write ### Algorithms - FCFS - fair but ineffcient - fifo - SSTF - starvation possible - select minimum seek time - **usual algorithm** - SCAN - elevator algorithm - possibility of double time - double direction - C-SCAN - one direction since disk is round - **usual algorithm **- provide uniform wait time - LOOK - one direction until services all requests → reverse - C-LOOK - variant of C-SCAN - only goes as far as the last request in each direction

e6e60cca761147059a1bf7717616c398

Disk layout

Disk Management Notion

Nov 5, 2019

Alan Jo

Seong-lae Cho

May 19, 2021

cpu can't directly control disk so kernel has in-memory sturcture ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2Fcf347120-7912-4dae-ae82-0b858e27757f%2Ftemp.jpg?table=block&id=cdd3a113-0151-4967-853c-5c59b1663de5&cache=v2) ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F8ba5baf0-04e8-425e-8fef-ec74cb81ac66%2FUntitled.png?table=block&id=79a04b2e-946f-4b24-93db-68bf497c9463&cache=v2) ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F7338d4d4-8caa-4323-bf34-c27267abf197%2FUntitled.png?table=block&id=74744e50-5351-46a0-b807-98e79499f478&cache=v2) ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F05994a8d-f0f0-4362-8359-d7984a89eab2%2FUntitled.png?table=block&id=eb9b0dd1-387a-4de2-ba9f-8f47759f634d&cache=v2)

5091a147a01a4ff2986c9d80e6aad326

Disk management

Disk Management Notion

May 19, 2021

Alan Jo

May 19, 2021

[Disk partitioning](https://texonom.com/disk-partitioning-129115f572724f5cbc570c7517151187) [Disk bad sector](https://texonom.com/disk-bad-sector-5cc7aac350854f0885cfa60f62bee9ce) [Disk Swap Space](https://texonom.com/disk-swap-space-5bf2726b42c14bf1abeeb90018408163)

2db57b8f7837407088208b308e51bbec

Disk Performance

Disk Management Notion

May 19, 2021

Alan Jo

May 19, 2021

### Performance - 1. cost, 2. speed & bandwidth, 3. reliability optical disk - permanent data storage → cloud in temporary ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2Fc5563700-295a-4a9d-8f1e-604d5a19d691%2FUntitled.png?table=block&id=8e30c91b-677f-420a-93ea-e7e39f363dd2&cache=v2) ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2Fda675d40-dc07-49e5-b1a5-5505aeed791b%2FUntitled.png?table=block&id=dd0f8f2c-c868-4f57-b9b1-0520b0e86759&cache=v2)

28e569d80cbe4c3b9e68857ac92550ea

RAID

Disk Management Notion

Nov 5, 2019

Alan Jo

Seong-lae Cho

Jun 7, 2021

## redundant array of inexpensive disks [Hardware RAID](https://texonom.com/hardware-raid-861083a0395649d18fa43e8b5dec9f51) ### RAID level - originally 5 level [RAID 0](https://texonom.com/raid-0-5b52502a6406464689c2bf557f5d3b47) [RAID 1](https://texonom.com/raid-1-b4b2a03a0193424e8fa40b4ee7488be7) [RAID 2](https://texonom.com/raid-2-1b24579f29674f05a13c00b5075f2739) [RAID 3 ](https://texonom.com/raid-3-4cc5f5d7b3c84145825a3f9bc4975ec9) [RAID 4](https://texonom.com/raid-4-f4db629dc21d4e58a8ad908a83be049f) [RAID 5](https://texonom.com/raid-5-afdf3a4add0a4374958c915fea1f83ef) Usual tech of server and NAS and data center ### Two design goals of various RAIDs 1. increased data reliabilitymirroring data 2. increased input/output performance trade-off between reliability and performance > even parity, odd parity for detect block failure > Striping : Disk striping - look at the diagrams which are like stripes

4b50ffdd255d430ca76f393e8d1ebe25

Tertiary storage devices

Disk Management Notion

Nov 5, 2019

Alan Jo

Seong-lae Cho

May 19, 2021

ex. usb, sd card, external disk [HSM](https://texonom.com/hsm-95b2b5fc37ca4982aed9c39ffe45ecb2)

def6411e9adf4220811b65706895e6e3

Disk bad sector

Disk management

null

[Disk bad sector remapping](https://texonom.com/disk-bad-sector-remapping-f18741071ab949c29a01c238403a0630)

5cc7aac350854f0885cfa60f62bee9ce

Disk partitioning

Disk management

null

129115f572724f5cbc570c7517151187

Disk Swap Space

Disk management

null

temporal image space of main memory ### Swap File approach ### Swap partition Approach - Swapping By File system read but slow because of using VFS but do not need partition - or do not need file system just image dump to this sector > rule of the thumb - twice of main memory

5bf2726b42c14bf1abeeb90018408163

Disk bad sector remapping

Disk bad sector

null

f18741071ab949c29a01c238403a0630

Hardware RAID

RAID

null

하드웨어 디바이스를 통해 RAID를 수행하는 경우 운영체제와 파일시스템은 RAID 메커니즘을 인식하지 못함 전용 하드웨어인 RAID card 자체에서 오류가 발생하는 경우 전체 디스크 array가 무용지물이 되는 경우도 > [네이버 클라우드 플랫폼 (NAVER Cloud Platform) : 네이버 블로그](https://blog.naver.com/n_cloudplatform/222384170516)

861083a0395649d18fa43e8b5dec9f51

RAID 0

RAID

null

![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F82335579-cbcb-400a-8260-2f4bcbb2ec06%2Ftemp.jpg?table=block&id=f24a948f-aad6-45fc-bd58-2d347fbc8e6c&cache=v2) - RAID 0 - striped disks

5b52502a6406464689c2bf557f5d3b47

RAID 1

RAID

null

![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F82335579-cbcb-400a-8260-2f4bcbb2ec06%2Ftemp.jpg?table=block&id=49941f40-28e1-47fd-9f36-f7f78bd839f9&cache=v2) - RAID 1 - disk mirroring

b4b2a03a0193424e8fa40b4ee7488be7

RAID 2

RAID

null

1b24579f29674f05a13c00b5075f2739

RAID 3

RAID

null

![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2Faafcb581-cf0c-4683-9ed0-c28bf5c9f412%2Ftemp.jpg?table=block&id=48020019-2481-4287-8a7e-3d635536caf4&cache=v2) - RAID 3 - Striped set with dedicated parityidea of redundant disk to increase reliability - many disk decrease reliabilitybut update parity block overhead and bottle neck to parity disk

4cc5f5d7b3c84145825a3f9bc4975ec9

RAID 4

RAID

null

f4db629dc21d4e58a8ad908a83be049f

RAID 5

RAID

null

![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2Ffeb998d5-2d36-4dc7-bf82-eb682687d18e%2Ftemp.jpg?table=block&id=a2fc4fbf-b4bb-419f-b0a9-2b231d62e11d&cache=v2) - RAID 5 - Striped set with distributed parity

afdf3a4add0a4374958c915fea1f83ef

HSM

Tertiary storage devices

null

95b2b5fc37ca4982aed9c39ffe45ecb2

Memory Address Changing

Memory Management Notion

Nov 5, 2019

Alan Jo

Seong-lae Cho

May 19, 2021

- **Physical address space** - Starting at address 0, going to address MAXsys - **Logical address space** - Starting at address 0, going to address MAXprog **CPU(code) execute instructions by logical address** > **relative address** is an example of logical addres**s** ### Address Generation ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2Fd35ce2bf-0ec6-44eb-a128-e63e6089a756%2FUntitled.png?table=block&id=2d6934fd-f96f-4c0d-a731-09f64ece4df1&cache=v2) until linking, code has relative address and after loading to memory, it has physical address → Hardware for Address Translation (For adequate performance) ← dynamically “relocated” - **Base (also called Relocation) Register **- used for validation in CPU (either limit R) - **Limit Register -** size of program ## Linking combining various pieces of code and data into a single file that can be loaded into memory <details><summary> Linker’s tasks</summary> 1. **Symbol resolution** : each **global symbol** in an object file is bound to a **unique definition** 2. **Relocation** : the physical **memory address for each symbol is determined **and where **references **to those objects are modified **either** </details> <details><summary> kind of linker</summary> 1. static linker(`linkage editor’) - Object modules combine while Compile time 2. dynamic linker - Useful with language libraries (no duplicated copies) - done in Load time or Run time </details> ### Shared library - Static libraries disadvantages - Potential for duplicating- **ex. libc.a** - automatically occur when executable - also occur by user code dlopen() - routines can be shared by multiple processes **ex. ****[libc.so](http://libc.so)**

53916e26496843d6bf5a26e6185ebe8d

Memory Management History

Memory Management Notion

May 19, 2021

Alan Jo

May 19, 2021

# Backgrounds Multi programming by time sharing Need multiple processes in memory at once (number of I/O + CPU) **So we need protected updating** **Memory management = **The task that is carried out by the “**OS and hardware(MMU)**” to accommodate multiple processes and **fully use memory resources **to maximize performance **Requirement - **Each process should have a logically contiguous space → virtual memory, make user do not consider memory while make program ### **Goal - protect, performance(fully use) and sharing**

baac71af6c7c4281894e15349bde6777

Memory Management Method

Memory Management Notion

May 19, 2021

Alan Jo

May 19, 2021

### Memory Management Methods |Title| |:-:| |[Virtual Memory](https://texonom.com/virtual-memory-07bf10ba636b42e9818c009b17f893f0)| |[Paging](https://texonom.com/paging-0571138c29914318a5f6a717cc2b36c0)| |[Dynamic Partitioning](https://texonom.com/dynamic-partitioning-ec4c829bbfec46d28f441044264eb5a7)| |[Fixed partitioning](https://texonom.com/fixed-partitioning-068a5eb6e57b4146926dc02fc828b649)| |[Segmentation](https://texonom.com/segmentation-9927cad8b8bf4366abdc591e533e1060)|

040223391bea4c148bcfee4f4064572f

Dynamic Partitioning

Memory Management Methods

Nov 5, 2019

Alan Jo

Seong-lae Cho

May 19, 2021

- make partition when process input come - Each process is allocated exactly as much memory as it requires - Has external fragment (hole is external with no internal) - Use compaction to shift processes to free memory by external (less use by i/o problem) > **Compaction is possible only if relocation is dynamic, and is done at execution time** > I/O problem - Do I/O only into Operating System buffers <details><summary> Possible algorithms of placement</summary> 1. Best-fit - The fragment as small as possible 2. First-fit 3. Next-fit: choose first hole from last placement 4. Worst-fit: allocate the largest hole </details> > **Overlays** - choose memory layout by program call graph to efficient memory distribute by function module code

ec4c829bbfec46d28f441044264eb5a7

Fixed partitioning

Memory Management Methods

Nov 5, 2019

Alan Jo

Seong-lae Cho

May 19, 2021

Partitions can be of equal or unequal sizes set by hardware - has internal fragmentation (Unequal-size partitions lessens these) - A program may be too large → design the program with overlays ### queue(s) fixed partitioning Placement algorithm with “unequal size” partitions use of multiple queues(minimize internal fragmentation) → some queue will empty → use of single queue (increase internal fragmentation) > internal means internal of partition

068a5eb6e57b4146926dc02fc828b649

Paging

Memory Management Methods

Nov 5, 2019

Alan Jo

Seong-lae Cho

May 19, 2021

### Physical address space of a process can be non-contiguous (Pages are contiguous) - Keep track of all free frames - internal fragment - Set up a page table to translate logical to physical addresses - A virtual (logical) address is a pair (p, o) - Physical memory address is a pair (f, o) - Not all pages are mapped to frames **내부 단편화 O, 외부 단편화 X** **고정된 영역인 페이지(page)로 분할** ### Paging Notion |Title| |:-:| |[Page Replacement Algorithm](https://texonom.com/page-replacement-algorithm-3721241bd6b84fd9b55960f82ecf1fbe)| |[Paging Table](https://texonom.com/paging-table-b368ef8543cf4628b837232f4fd7fe7c)| |[Associative Register](https://texonom.com/associative-register-538b0ad40e84419795da36b76959786f)| |[Paging Protection ](https://texonom.com/paging-protection-b0940a815ff94260bcae736355499b4b)| |[Hierarchical paging](https://texonom.com/hierarchical-paging-2b45d0a7145b4ae98735812f9f771786)| |[Segmentation Paging](https://texonom.com/segmentation-paging-dfb1e2f233f1401daf3f8dc74466a57e)| |[Buddy System](https://texonom.com/buddy-system-b83b6f21d398442a84ff9c4fcd2ce295)| ### Paging Usage |Title| |:-:| |[Paging Example](https://texonom.com/paging-example-35326cf37c694b23ab5bab18de683b9f)| > [세그먼테이션(segmentation) VS 페이징(paging)](https://blog.naver.com/PostView.nhn?blogId=qbxlvnf11&logNo=221368476938&parentCategoryNo=&categoryNo=&viewDate=&isShowPopularPosts=false&from=postView)

0571138c29914318a5f6a717cc2b36c0

Segmentation

Memory Management Methods

Nov 5, 2019

Alan Jo

Seong-lae Cho

May 19, 2021

### more classical method before paging but used in virtual address image **가변적인 영역인 세그먼트(segment)로** **내부 단편화 X, 외부 단편화 O** - like paging - it has segment table also in each process Logical address consists of a two tuple - <segment-number, offset> Segment-table base register (STBR), Segment-table length register (STLR) exist - **has segment number s(identify segment) and d index(of process entry)- Maps two-dimensional physical addresses**each entry has base and limit ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2Fa58476a4-a0c8-4077-aa61-4aa199746880%2FUntitled.png?table=block&id=2c19b285-1a31-4664-8470-b281f9517a36&cache=v2) Has also relocation, sharing, allocation issue for fragmentation ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F74859a9e-2904-49ab-80d3-450c8e24a75c%2FUntitled.png?table=block&id=4927863d-e58e-41d0-9615-2fa88e16f7fa&cache=v2) > [세그먼테이션(segmentation) VS 페이징(paging)](https://blog.naver.com/PostView.nhn?blogId=qbxlvnf11&logNo=221368476938&parentCategoryNo=&categoryNo=&viewDate=&isShowPopularPosts=false&from=postView)

9927cad8b8bf4366abdc591e533e1060

Virtual Memory

Memory Management Methods

Nov 5, 2019

Alan Jo

Seong-lae Cho

Jan 18, 2023

## Demand paging = virtual memory ### Motive : In any hardware, same code(address) should have same data (memory management by maintaining logical semantic) virtual address: logical address **Concept : separating of user logical memory from physical memory** > - Symbolic Address : access variable by name > - address binding : variable mapping to address and physical memory when compile **Insight : program do not need all of their code and data at once, so we can separate (but of course have overhead)** > - Implementation : demand paging **effect : VM enables program to execute without requiring their entire address space to be resident in physical memory. isolate each process's memory, enable share files and address spaces.** ### Virtual Memory Notion |Title| |:-:| |[Page Table](https://texonom.com/page-table-ffbce6c10d9a4a8993bd0d809f4a4249)| |[Segmentation Fault](https://texonom.com/segmentation-fault-b1008afd3e3a4117a92c978ba3dc0c14)| |[Page Replacement](https://texonom.com/page-replacement-936ec615a0554950bc1ab233fb67434f)| |[page size](https://texonom.com/page-size-c50803e5e52b492c847de02404f0be36)| |[TLB reach](https://texonom.com/tlb-reach-afd1dceecb9542c086eab1d89b938824)| |[Swapping](https://texonom.com/swapping-98903f7a115e48808b17fc418468c8e3)| |[Page fault handler](https://texonom.com/page-fault-handler-b4d166a24a7740a9bc153f80701a03bc)| |[Virtual Memory Thrashing](https://texonom.com/virtual-memory-thrashing-f57fa2a2a76244278cca8ad66badd63f)| |[i/o interlock](https://texonom.com/io-interlock-458727109a584d01b1087036dd6922bc)| |[Pages Arguments](https://texonom.com/pages-arguments-287308a9ec104a0eb0c4e01b157c8831)| |[TLB (Translation Lookaside Buffer)](https://texonom.com/tlb-translation-lookaside-buffer-fd5ffc7e77c44f57a744915cc426c27e)| ### Virtual Memory Notion Usages |Title| |:-:| |[Linux Paging](https://texonom.com/linux-paging-dddeb1bb20e64c67bf7d38ae744e625c)| |[COW(copy-on-write)](https://texonom.com/cowcopy-on-write-b41107af21f04feb923f02eeaf677017)| |[Memory-Mapped Files](https://texonom.com/memory-mapped-files-09442c8a229e475083829ddf52eca697)| > [8강. 메모리 관리](https://marlinbar.tistory.com/68?category=300896)

07bf10ba636b42e9818c009b17f893f0

Associative Register

Paging Notion

Nov 5, 2019

Alan Jo

Seong-lae Cho

May 19, 2021

parallel search - if associative register exist <details><summary> Address translation (A´, A´´) </summary> – If A´ is in associative register, get frame number out – Otherwise, get frame number from page table in memory </details> ### Effective Access time Associative lookup = `math: \epsilon `time unit - related to hardware performanc Assume memory cycle time is 1 microsecond Hit ratio (`math: \alpha`) - related to algorithm $$EAT = (1 + \epsilon)\alpha + (2 + \epsilon)(1 - \epsilon) \\ = 2 + \epsilon - \alpha$$

538b0ad40e84419795da36b76959786f

Buddy System

Paging Notion

Nov 5, 2019

Alan Jo

Seong-lae Cho

May 19, 2021

- used with virtual memory system either - combination of fixed and dynamic Compromise to overcome disadvantages of both fixed and dynamic partitioning schemes <details><summary> **Alogorithm**</summary> 1. Start with entire block 2. When request of size S is made > ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2Fa14d3c04-3016-4f0e-a409-5db1a488c95a%2FUntitled.png?table=block&id=1372aeb8-b364-4946-8e46-1cc0c6ac5fc5&cache=v2) 3. this process is repeated until the smallest block greater or equal to S is generated. 4. Two buddies are coalesced whenever both of them become de-allocated. > Given the address and size of a block, the address of buddy is automatically determined. The address of a block and its buddy differ in exactly one bit position > **binary branch framework and buddy(share one parent) is has different bit of final address** </details> ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F95bed60f-0f6b-4160-993d-7960c4d49888%2FUntitled.png?table=block&id=f3fd7300-747a-4204-b62f-8b947fe5c206&cache=v2) > modified form is used in UNIX SVR4 (and also in Linux) for kernel memory allocation (KMA) > Smallest size of block is page ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F4085edf3-c409-4719-b832-70c192b29e91%2FUntitled.png?table=block&id=5c4d8207-1c56-4bf2-9233-3a534cbe810d&cache=v2)

b83b6f21d398442a84ff9c4fcd2ce295

Hierarchical paging

Paging Notion

Nov 5, 2019

Alan Jo

Seong-lae Cho

May 19, 2021

– A simple technique is a two-level page table – Break up the logical address space into multiple page table Has page offset either but **has multi page number in logical address** ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F8138952b-d75a-443f-92a3-5dfbabd137b2%2FUntitled.png?table=block&id=5b07a593-5c04-4e1e-a521-806dc03a80fe&cache=v2) last page number indicate page of page table,other is root or outer page table

2b45d0a7145b4ae98735812f9f771786

Page Replacement Algorithm

Paging Notion

May 19, 2021

Alan Jo

May 19, 2021

### Page Replacement Algorithms |Title| |:-:| |[LRU Algorithm](https://texonom.com/lru-algorithm-1f181112165f4df8b5a5cdea967d65c1)| |[Clock Algorithm](https://texonom.com/clock-algorithm-80f852e82a0548f5856f09baf11f6cf1)| |[Second Chance Algorithm](https://texonom.com/second-chance-algorithm-afdfb27ecaac4854902ba38eec679e41)| |[FIFO Algorithm](https://texonom.com/fifo-algorithm-80eb13636a8f4edc8861c04033c9d637)| |[LFU Algorithm](https://texonom.com/lfu-algorithm-db665029bdb84a42a98099c471ec744b)| > [우주를 위해 : 네이버 블로그](https://blog.naver.com/oidoman/90097323379) > [LRU 알고리즘 (Least Recently Used Algorithm)](https://j2wooooo.tistory.com/121)

3721241bd6b84fd9b55960f82ecf1fbe

Paging Protection

Paging Notion

Nov 5, 2019

Alan Jo

Seong-lae Cho

May 19, 2021

**Memory protection bits** - Implemented by associating protection bit with each page - e.g. read-only, read-write, in page table **valid/invalid bit** - changed by other page in same frame ### Sharing is easy - just indicate same frame and make it both valid

b0940a815ff94260bcae736355499b4b

Paging Table

Paging Notion

May 19, 2021

Alan Jo

May 19, 2021

### Paging Tables |Title| |:-:| |[Hashed page table](https://texonom.com/hashed-page-table-5503e136b5d94731911edb2a21ccd4c9)| |[Inverted page table](https://texonom.com/inverted-page-table-25d3365925cf4aea9b9eb834c1c6bbad)| ### One table (mean one image) for each process: part of process’s state <details><summary> Contents</summary> 1. Flags: valid/invalid (also called resident) bit, dirty bit, reference (also called clock or used) bit used for clock page change algorithm 2. Page frame number </details> Page table is kept in main memory - Page-table base register (PTBR) points to the page table of each process - Page-table length register (PTLR) indicates size of the page table of each process > use tlb to lessen memory access overhead to find page page table size is pretty big → –** Only need to map the portion of the address space actually being used** (tiny fraction of entire address space) Approach Methods > – Multi-level(Hierarchical) paging, Hashed page table, Inverted page table

b368ef8543cf4628b837232f4fd7fe7c

Segmentation Paging

Paging Notion

Nov 5, 2019

Alan Jo

Seong-lae Cho

May 19, 2021

- **Use segments to manage logically related units** - **Use pages to partition segments into fixed size chunks **- first go to segment table (protection easy) > Need to allocate page table entries only for those pieces of the segments that have themselves been allocated. > ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2Fd370d686-0d0a-4b36-9304-1c52bb619bfa%2FUntitled.png?table=block&id=33d0e559-e0c3-4c22-9dcb-38bbc8eeb185&cache=v2) > ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F5c5ffdbb-d91c-4b65-b989-fde20a0389cf%2FUntitled.png?table=block&id=69399c6b-6efa-4b2e-a6bc-3665fc417f01&cache=v2) **Logical (virtual) address → Linear address → Physical address (intel 32 architecture)** ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2Fd6130a43-2117-4675-acf7-49442a5ed895%2FUntitled.png?table=block&id=8bd0a58e-360f-499a-b136-4a2bd00ce504&cache=v2) Segmentation: logical partition of memory - Segment Selector bits exist (segmentation register store segment selector bits usually) ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2Ff8645193-e5ef-4450-a3af-26da99b9b36c%2FUntitled.png?table=block&id=2b2edf19-d47f-4964-9766-a221553cf019&cache=v2) > Offset means relative address

dfb1e2f233f1401daf3f8dc74466a57e

Clock Algorithm

Page Replacement Algorithms

May 19, 2021

Alan Jo

May 19, 2021

second-chance보다 더 효율적인 FIFO버전의 알고리즘 queue의 페이지를 이동시키지 않는다. clock 알고리즘은 순환형 리스트를 유지하고, 가장 오래된 페이지를 가리키는 hand(iterator)를 가지고 있다 페이지 폴트가 발생했고 빈 프레임이 없다면, hand가 가리키고 있는 페이지의 참조비트를 검사 만약 참조비트가 클리어 상태라면 hand가 가리키고 있는 페이지가 새로운 페이지로 사용 그렇지 않으면 참조비트를 클리어시킨다. 그 다음 hand는 증가하고, 페이지가 교체될 때까지 이 과정이 반복 > [우주를 위해 : 네이버 블로그](https://blog.naver.com/oidoman/90097323379)

80f852e82a0548f5856f09baf11f6cf1

FIFO Algorithm

Page Replacement Algorithms

May 19, 2021

Alan Jo

May 19, 2021

**FIFO** : 페이지가 주기억장치에 적재된 시간을 기준으로 교체될 페이지를 선정하는 기법 단점 : 중요한 페이지가 오래 있었다는 이유만으로 교체되는 불합리. 가장 오래 있었던 페이지는 앞으로 계속 사용될 가능성이 있음. [Belady's Anomaly](https://texonom.com/beladys-anomaly-d5dfd0e9467f4aa1a7ba5d584c4c6d0c)

80eb13636a8f4edc8861c04033c9d637

LFU Algorithm

Page Replacement Algorithms

May 19, 2021

Alan Jo

May 19, 2021

**LFU** : 가장 적은 횟수를 참조하는 페이지를 교체 단점 : 참조될 가능성이 많음에도 불구하고 횟수에 의한 방법이므로 최근에 사용된 프로그램을 교체시킬 가능성이 있고, 해당 횟수를 증가시키므로 오버헤드 발생

db665029bdb84a42a98099c471ec744b

LRU Algorithm

Page Replacement Algorithms

May 19, 2021

Alan Jo

May 19, 2021

## Least Recently Used Algorithm 가장 오랫동안 참조되지 않은 페이지를 교체

1f181112165f4df8b5a5cdea967d65c1

Second Chance Algorithm

Page Replacement Algorithms

May 19, 2021

Alan Jo

May 19, 2021

FIFO의 변형된 형태의 알고리즘 FIFO와 같이 제거될 페이지를 queue의 첫번째 요소로부터 찾는다 그러나 즉시 페이지를 제거하는 것이 아니라, 그 페이지의 참조비트가 세팅되어 있는지를 체크한다 세팅되어 있지 않으면 제거하는 것이고, 그렇지 않으면 참조비트를 클리어해서 queue의 맨 뒤로 이동시 > [우주를 위해 : 네이버 블로그](https://blog.naver.com/oidoman/90097323379)

afdfb27ecaac4854902ba38eec679e41

Belady's Anomaly

FIFO Algorithm

null

FIFO(First In First Out)에서, 원래 페이지 프레임의 개수를 늘리면 page fault발생이 감소 해야 하나, 오히려 늘어나는 경우가 발생 > [벨레이디의 모순 (Belady's Anomaly)](https://m.blog.naver.com/cookatrice/50037414550)

d5dfd0e9467f4aa1a7ba5d584c4c6d0c

Hashed page table

Paging Tables

Nov 5, 2019

Alan Jo

Seong-lae Cho

May 19, 2021

- The virtual page number is hashed into a page table - This page table contains a chain of elements hashing to the same location - need hash function and page table like hash table ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F054a272e-ba5a-447a-b1fa-9f432abbc97b%2FUntitled.png?table=block&id=cdb4b8cb-54a6-4019-81e5-e68ca1ad2580&cache=v2) Virtual page numbers are compared in this chain searching for a match. If a match is found, the corresponding physical frame is extracted (fast)

5503e136b5d94731911edb2a21ccd4c9

Inverted page table

Paging Tables

Nov 5, 2019

Alan Jo

Seong-lae Cho

May 19, 2021

### 역 페이지 테이블 One entry for each real page of memory (frame) - Only one page table in the system - Entry consists of the virtual address of the page stored in that real memory location, with information about the process that owns that page - by pid, p → get i by hash table for limit one either ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F5dfe95bd-9a3d-4589-9252-dd6a7c0c0213%2FUntitled.png?table=block&id=e6e67336-8bae-4204-b6aa-13a3029a7c15&cache=v2) > [역 페이지 테이블 (Inverted Page Table) > 도리의 디지털라이프](http://blog.skby.net/%EC%97%AD-%ED%8E%98%EC%9D%B4%EC%A7%80-%ED%85%8C%EC%9D%B4%EB%B8%94-inverted-page-table/)

25d3365925cf4aea9b9eb834c1c6bbad

Paging Example

Paging Usage

May 19, 2021

Alan Jo

May 19, 2021

# Paging in Hardware (Intel x86) – **physical addresses of page tables (CR3 register store) - two level in intel (+ page directory)** ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2Fce48bc62-3c64-4981-9923-613303ea798a%2FUntitled.png?table=block&id=14d204a8-6dd8-4444-8190-33197645aa4b&cache=v2) Enable by setting PG bit in CR0 register 4kb so 20 upper bits to find page table ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2Ff3239cab-044c-4eae-ab4b-7782ae6ba8bf%2FUntitled.png?table=block&id=217f7801-2b3e-4b75-a709-0b6d37ce1c00&cache=v2) ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F11c86ab6-4414-47d7-a314-c6c3b4610e28%2FUntitled.png?table=block&id=79f5e569-5327-4fb8-9bac-bff87c58639c&cache=v2) ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F4bdc9f84-8bc1-4229-ac66-ebbc796c9ba8%2FUntitled.png?table=block&id=c4967232-f546-42b6-b243-c52905d4b111&cache=v2) > Linux support Four-level paging for future ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2Fb10eea4f-ecbb-44d7-8b3f-dacb0d1b89e7%2FUntitled.png?table=block&id=b3f394dc-0641-4f29-b741-7f20dfb188f7&cache=v2) ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F77e637c7-abbb-490c-86a4-5238dc928c7b%2FUntitled.png?table=block&id=03cdfda0-1c14-4877-8aeb-b01535c0e193&cache=v2) ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2Ff031c764-89ce-4d2e-a42f-037c38dc9cbe%2FUntitled.png?table=block&id=7097a2b7-c336-4177-aa5e-97a8c05f3632&cache=v2)

35326cf37c694b23ab5bab18de683b9f

i/o interlock

Virtual Memory Notion

Nov 5, 2019

Alan Jo

Seong-lae Cho

Jan 18, 2023

- 커널로 문제해결 - Lock pages in physical memory using lock bits.

458727109a584d01b1087036dd6922bc

Page fault handler

Virtual Memory Notion

Nov 5, 2019

Alan Jo

Seong-lae Cho

Jan 18, 2023

### It is kind of interrupt if required PTE (entry) has invalid bit in page table than cause an exception to throw so go to OS(kernel mode), bring in missing page into physical frame, update PTE and restart instruction. -**Page fault service time: **service the page-fault interrupt + Read in the page + Restart the process - **Effective access time: **memory access time * (1 - p) + disk access time * p (p = probability of page fault) **degradation: **effective access time / memory access time (low p than low degradation) > **Locality**: dependent on page replacement policy > page change occur by not only by PFH, since predict page usage by machine learning

b4d166a24a7740a9bc153f80701a03bc

Page Replacement

Virtual Memory Notion

Nov 5, 2019

Alan Jo

Seong-lae Cho

Jan 18, 2023

> **Belady's Proof (Oracle solution: ideal)**: Evicting the page that will not be used for the longest period of time minimizes the number of page faults # Counting Algorithms - LFU(Least frequently used)but do not reflect time variable*recent used victim - MFU(Most Frequently Used)consider recent used # Timing algorithms - FIFOSimple: single pointer sufficesBelady's Anomaly: more frame do not improve fifo algorithm performance - LRU (Least recently used)by counters or stack ### but lru is complicate ### LRU Approximation (real use) - Sampled LRU (hardware support)reference bit - can be cleared when full - Second-Chance AlgorithmLRU clock(Secondd chance) - clock bitclock hand - PTE reference - often use modified bit(dirty bit) either > Evaluate algorithm usually by running it on a particular string of memory references (reference string) and computing the number of page faults on that string **Evict picked victim page** Page fault service time – Service the page-fault interrupt + Read in the page + Restart the process ### Allocation of Frames each process need minimum number of pages - Must provide enough frames to hold all the different pages that any single instruction can reference - local replacement - can not steal other process's frame - Fixed(Equal) allocation - each algorithm to process - global replacement - can steal other process's frame - Priority allocation > Proportional allocation : related to process size set frame number > Priority allocation : set process priority # Page replacement Examples 1. windows NTdemand paging with clustering(gist)process has working-set min, working-set maximum and if exceed then automatic working-set trimming process perform to restore free memory 2. solaris 2 > ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2Ff66a2160-4cbc-4f7c-95f5-593519e3a76b%2FUntitled.png?table=block&id=21bfeffc-ef6d-43a1-b507-7f31bce81cbd&cache=v2)

936ec615a0554950bc1ab233fb67434f

page size

Virtual Memory Notion

Nov 5, 2019

Alan Jo

Seong-lae Cho

Jan 18, 2023

- Provide Multiple Page Sizeshardware page size various so we can choose by softwarelarge: small page table, small page faultsmall: less handling overhead, small internal fragment

c50803e5e52b492c847de02404f0be36

Page Table

Virtual Memory Notion

Nov 5, 2019

Alan Jo

Seong-lae Cho

Jan 18, 2023

Each Process has Page Table, Page table entry(PTE) has valid, access bit ### Page Table Arguments 1. V = {0 ,1 , ..., N-1}: virtual address space 2. P = {0, 1, ..., M-1}: physical address space 3. N > M: difference that not comparable 4. MAP: V → P U {null} address mapping function 5. null if in Disk (if valid is 0): page fault ### Page Table Parameters log upper = lower 6. P = Page size (bytes): typically KBs 7. N = Virtual Address limit 8. M = Physical address limit <details><summary> virtual address = virtual page number(VPN) bit + page offset</summary> VPN acts page index in page table </details> physical address = physical page number bit + page offset address translation change previous one only (32 bit to other) > compare start by page table base register page table need protection and cache do not need because of address translation

ffbce6c10d9a4a8993bd0d809f4a4249

Pages Arguments

Virtual Memory Notion

Nov 5, 2019

Alan Jo

Seong-lae Cho

Jan 18, 2023

Virtual page and Physical page both exists, Page Table connect VP to PP - P = total number of virtual pages - N = total number of Physical pages 1. PP can be placed in Disk (Storage) 2. Different VP can point out same PP (ex. library) > if disk read occur than DMA(direct memory access) transfer and then by I/O controller there will be interrupt processor

287308a9ec104a0eb0c4e01b157c8831

Segmentation Fault

Virtual Memory Notion

Jun 1, 2021

Alan Jo

Jan 18, 2023

[Segmentation](https://texonom.com/segmentation-9927cad8b8bf4366abdc591e533e1060)

## Segfault OS에 메모리 액세스 위반을 알리기 위해 메모리 보호 기능이있는 하드웨어에서 발생하는 오류 customer handler를 설치하여 사용하지 않으면 Linux의 커널에서 동작하고 있는 process signal 처리 segfault는 일반적으로 프로세스를 종료하고 적절한 ulimit 설정을 갖춘 코어 덤프를 생성 [Segfault RSP](https://texonom.com/segfault-rsp-9fa0f64a2de94e26906b7689c0590fd4) [Segfault RIP](https://texonom.com/segfault-rip-364a0188edf74a3e9f535a5f733bc996) 오류 값은 페이지 오류 오류 코드 비트의 비트 마스크 from `arch/x86/mm/fault.c` > [What is Segfault Message?](https://hslee77.github.io/linux,cloud/2021/05/07/LinuxSegfault-Event.html)

b1008afd3e3a4117a92c978ba3dc0c14

Swapping

Virtual Memory Notion

Nov 5, 2019

Alan Jo

Seong-lae Cho

Jan 18, 2023

A process can be swapped temporarily out of memory to a **backing store - **and then brought back into memory for continued execution Backing store - Fast disk that must provide direct access to these memory images bring a page into memory only when it is needed (page-level **swapping**) - OS do movement of pages between memory/disks - evicted pages go to disk if page is dirty to a swap file

98903f7a115e48808b17fc418468c8e3

TLB reach

Virtual Memory Notion

Nov 5, 2019

Alan Jo

Seong-lae Cho

Jan 18, 2023

- TLB Reach = (RLB Size) * (Page Size)가변 페이지 사이즈 이용해서 커널한테는 큰 페이지

afd1dceecb9542c086eab1d89b938824

TLB (Translation Lookaside Buffer)

Virtual Memory Notion

Nov 5, 2019

Alan Jo

Seong-lae Cho

Jan 18, 2023

usually fully associative cache since page table is in memory ### bits: valid + dirty(since it is cache and if write back) + tag(process ID + virtual page number) + physical page number + page offset

fd5ffc7e77c44f57a744915cc426c27e

Virtual Memory Thrashing

Virtual Memory Notion

Nov 5, 2019

Alan Jo

Seong-lae Cho

Jan 18, 2023

<details><summary> Since lack of free frame(low RAM), process > physical memory then page fault rate will be very high</summary> ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F42e6be14-9a28-4c41-9344-709852b69ee1%2Ftemp.jpg?table=block&id=7f7afe8b-044d-458c-af60-912fb3450ba1&cache=v2) </details> → os use all time to swapping Solution Method two - working set, ppf > ### 1. Working set, locality set + others for continity = working set ***Working Set***: set of actively used pages `math: \Delta` : working-set window : fixed number of page references (ex. 10,000 instruction) > ex. WS(t1) = {1,2,3,4,5} ,한꺼번에 반납, 대여 $$\Sigma WSSi = total demand frames$$ OS monitor direct D, and if D > total number of frames in physical them swap process ***then how to direct?: ***like sampled lru use ***reference bit*** and ***interval time*** increasing bit will make performance better 2^bit per interrupt ### 2. Page-fault Frequency Allocation Window's method directly measure and control page-fault rate establish acceptable page fault frequency - too high rate → increase number of frames - too low rate → decrease number of frames

f57fa2a2a76244278cca8ad66badd63f

Segfault RIP

Segmentation Fault

null

명령 포인터 레지스터 값

364a0188edf74a3e9f535a5f733bc996

Segfault RSP

Segmentation Fault

null

스택 포인터 레지스터 값

9fa0f64a2de94e26906b7689c0590fd4

COW(copy-on-write)

Virtual Memory Notion Usages

Nov 5, 2019

Alan Jo

Seong-lae Cho

Jan 18, 2023

like fork(do not need to copy code and data) we do resource sharing efficiently image copy on write time, parent and child processes to initially share the same If either process modifies a shared page, only then is the page copied. ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F024bcd14-9529-4600-bd90-9c52a217c55e%2FUntitled.png?table=block&id=10970c8f-03c3-4373-9c16-b2c1748a6a49&cache=v2)

b41107af21f04feb923f02eeaf677017

Linux Paging

Virtual Memory Notion Usages

Nov 5, 2019

Alan Jo

Seong-lae Cho

Jan 18, 2023

### four-level paging for future hardware which has many three level to mapping to hardware each process has its own Page Global Directory and its own set of page (upper, middle, null)tables ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F81e93678-3bfb-4c92-ad67-8febcbb65be7%2Ftemp.jpg?table=block&id=0c97af39-ebe2-4696-b377-c3c931e8f15f&cache=v2) ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F0d77a5e8-bea8-4ece-82bd-39352f2d73e0%2Ftemp.jpg?table=block&id=fad97902-3c20-4c54-a857-31c1235cc219&cache=v2) task struct data sturucture has mm pointer, mm is pointer which has image mm_struct data sturcture. that has mmap pointer. process's meta data sturcture vm araa struct pgd is page first table base address directory 1. read then page fault (segmentation fault) start end area 2. write code is fault so OS read commision 3. also fault because because r/w no physical in frame then replacement > ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2Fe3a42f71-e4d2-4057-97b8-80ace7bef2d1%2FUntitled.png?table=block&id=4dd69d90-4471-4c46-be35-7f144ff711d8&cache=v2) ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F56c9d36a-87cf-4f5b-b9ad-cbde6a4d621d%2Ftemp.jpg?table=block&id=67c9f0be-bd02-44ac-8eeb-48005baed1a4&cache=v2) **KMA: **dynamic memory allocation in head (3gb) by malloc. in kernel (1gb) memory allocationbecause kernel cannot use malloc - contiguous page frame allocator : alooc_pages() - noncontiguous page fram allocator : vmalloc() - memory object allocators : the slab alloctor- kmem_cache_alloc : congifuous allocation of kernel objects- kmalloc() : contiguous allocation of arbitrary length up to 128kb ## Page replacemnet in linux - **two LRU listsinactive_list - pages not known to be importantactive_list - pages known to be referenced frequently** - kernel swap daemon (a kernel thread) - more on page replacemnet mechanism > 64bit usually use 3 level paging

dddeb1bb20e64c67bf7d38ae744e625c

Memory-Mapped Files

Virtual Memory Notion Usages

Nov 5, 2019

Alan Jo

Seong-lae Cho

Jan 18, 2023

Memory-mapped file I/O allows file I/O to be treated as routine memory access by mapping a disk block to a page in memory - Uniform access for files and memory → less copying ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F26e99936-66b7-4614-ad26-1cc2b825c286%2FUntitled.png?table=block&id=02e0e36d-a529-43de-a147-3a10034222b8&cache=v2)

09442c8a229e475083829ddf52eca697

Garbage Collector

Memory Management Usage

Oct 29, 2020

Alan Jo

May 19, 2021

### Garbage Collector Notion |Title| |:-:| |[GC Compaction](https://texonom.com/gc-compaction-9179519ad2ec4ba48377ad941be50edf)| |[Generational GC](https://texonom.com/generational-gc-9e88cd3b020e417dbe4d0a231884d0c7)| |[Generational Theory](https://texonom.com/generational-theory-307b210f9661421580b4b22b331014c1)| ### Garbage Collectors |Title| |:-:| |[Static Garbage Collector](https://texonom.com/static-garbage-collector-eaac7f4e278741fe97fa212dd9037eba)| |[Dynamic Garbage Collector](https://texonom.com/dynamic-garbage-collector-7d88c0e0170b4c1c8a890b6e65932088)|

8af9f92e3ca842ee8ddd638e25647cd7

Memory Leak

Memory Management Usage

Oct 29, 2020

Alan Jo

May 19, 2021

1. Global Variable 2. Timer, Callback 3. Closer 4. Multi Reference > [자바스크립트와 V8 엔진의 메모리 관리 프로세스](https://medium.com/naver-place-dev/%EC%9E%90%EB%B0%94%EC%8A%A4%ED%81%AC%EB%A6%BD%ED%8A%B8%EC%99%80-v8-%EC%97%94%EC%A7%84%EC%9D%98-%EB%A9%94%EB%AA%A8%EB%A6%AC-%EA%B4%80%EB%A6%AC-%ED%94%84%EB%A1%9C%EC%84%B8%EC%8A%A4-f45091e696e1)

16c4f840b59e4d45a04770b03e582229

GC Compaction

Garbage Collector Notion

Apr 30, 2021

Alan Jo

Apr 30, 2021

[Dynamic Garbage Collector](https://texonom.com/dynamic-garbage-collector-7d88c0e0170b4c1c8a890b6e65932088)

![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2Ff1cd8946-653d-4fa4-b9a8-d9d16a199728%2FUntitled.png?table=block&id=697cd2a4-4784-415e-b7dc-5f1465129ff1&cache=v2) > [Go 언어의 GC에 대해 - LINE ENGINEERING](https://engineering.linecorp.com/ko/blog/go-gc/)

9179519ad2ec4ba48377ad941be50edf

Generational GC

Garbage Collector Notion

Apr 30, 2021

Alan Jo

Apr 30, 2021

[Generational Theory](https://texonom.com/generational-theory-307b210f9661421580b4b22b331014c1) [V8 GC](https://texonom.com/v8-gc-2630948f867a46069a56ccae5271f574) [JVM GC](https://texonom.com/jvm-gc-0023b0c7ffe644ed94a45f2a69fde1ea)

세대별 GC의 목적은 heap 내 객체를 수명(GC에서 살아 남은 횟수 등으로 나타냄)에 따라 분류하여 GC 효율을 향상시키는 것 - 객체 신규 할당 영역에선 GC를 자주 수행(Minor GC) - 해당 영역 GC에서 여러 번 살아 남은(가장 오래된) 객체는 승격(promote)시켜 GC 빈도가 낮은(Major GC) 영역으로 이동 Java8 HotSpot VM에서는 모든 컬렉터가 세대별 GC를 제공 [write barrier](https://texonom.com/write-barrier-73c82a9b955346429fee7fc504ab3172) > [Go 언어의 GC에 대해 - LINE ENGINEERING](https://engineering.linecorp.com/ko/blog/go-gc/)

9e88cd3b020e417dbe4d0a231884d0c7

Generational Theory

Garbage Collector Notion

Apr 30, 2021

Alan Jo

Apr 30, 2021

새로 할당된 객체는 대부분 일찍 죽는다는 가설

307b210f9661421580b4b22b331014c1

write barrier

Generational GC

null

세대별 GC는 GC를 수행하지 않을 때도 애플리케이션에 오버헤드가 발생한다는 단점 root에서 신규 세대를 참조하는 경로만 조사한 뒤 그 중 접근 불가능한 상태인 것을 회수하면 그렇다고 오래된 객체를 포함하여 heap 전체를 검사해 버리면 세대별 GC를 하는 의미가 없다 그래서 애플리케이션에서 참조를 대입하거나 rewrite할 때, 오래된 세대에서 신규 세대를 참조하는 경우 해당 참조를 별도로 기록하는 처리가 추가 이와 같은 참조의 ‘mutate’와 함께 부수적으로 진행해야 하는 처리를 ‘write barrier’라고 ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2F91b10137-4b01-4e06-8395-497fe3e92e6e%2FUntitled.png?table=block&id=65bf9e6d-3ddb-4cee-bf93-c23e71d43272&cache=v2) 세대별 GC를 사용하면 write barrier 오버헤드에 비해 얻을 수 있는 이점이 더 클 것으로 기대 > [Go 언어의 GC에 대해 - LINE ENGINEERING](https://engineering.linecorp.com/ko/blog/go-gc/)

73c82a9b955346429fee7fc504ab3172

Dynamic Garbage Collector

Garbage Collectors

Apr 30, 2021

Alan Jo

Apr 30, 2021

GC 수행 시 alive 상태의 객체를 heap 끝으로 재배치해서 heap을 압축 - 단편화 회피 - bump allocation을 통해 고속 메모리 할당 구현 가능 > - 객체가 heap 끝에 위치하므로 메모리 신규 할당 시 끝에서 바로 increment ![](https://www.notion.so/image/https%3A%2F%2Fs3-us-west-2.amazonaws.com%2Fsecure.notion-static.com%2Ffd98b5ba-7490-479b-a026-7bcf994a6328%2FUntitled.png?table=block&id=1080d55c-9308-438e-bc75-e6e490419519&cache=v2) ### Dynamic Garbage Collectors |Title| |:-:| > [Go 언어의 GC에 대해 - LINE ENGINEERING](https://engineering.linecorp.com/ko/blog/go-gc/)

7d88c0e0170b4c1c8a890b6e65932088

Static Garbage Collector

Garbage Collectors

Apr 30, 2021

Alan Jo

Apr 30, 2021

정적 유형의 GC는 GC별로 heap 내 객체를 재배치하지 않습니다 정적 유형 GC에서 메모리 할당과 해제를 반복하면 heap 단편화가 발생해 할당 성능이 악화되는 문제가 있다 단, 이는 메모리 할당자를 어떻게 구현하느냐에 따라 달라짐 ### Static Garbage Collectors |Title| |:-:| |[Tracing garbage collection](https://texonom.com/tracing-garbage-collection-19e036fe4c964d12be25ea3a974b69fd)| > [Go 언어의 GC에 대해 - LINE ENGINEERING](https://engineering.linecorp.com/ko/blog/go-gc/)

eaac7f4e278741fe97fa212dd9037eba

Tracing garbage collection

Static Garbage Collectors

Nov 9, 2020

Alan Jo

Apr 30, 2021

[Stop the world GC](https://texonom.com/stop-the-world-gc-5b16f1ca6ac3450bbf63e3bf56397538) [Mark & Sweep GC](https://texonom.com/mark-sweep-gc-b7bb682c52bc4be98043c847381fe483) [Reference Counting](https://texonom.com/reference-counting-a780282e427c436f8e12ca7f95173926) [Strong and weak references](https://texonom.com/strong-and-weak-references-7c3454cf4f1843d991bdc68439ec0091) > [Tracing garbage collection](https://en.wikipedia.org/wiki/Tracing_garbage_collection#Stop-the-world_vs._incremental_vs._concurrent)

19e036fe4c964d12be25ea3a974b69fd

Mark & Sweep GC

Tracing garbage collection

null

### 루트에서 접근 가능한 메모리는 active, 나머지는 기비지 [CMS Collector](https://texonom.com/cms-collector-5893234d493d4a9a8e14b4a392d44827) **active root 에 원치 않는 참조가 포함되어있다면, 필요 없는 메모리가 할당된 만큼 메모리 누수가 발생** 2단계 가비지 컬렉션 알고리즘 1. Mark: GC 루트 목록을 생성합니다. 자바스크립트의 경우 `window`객체가 루트로 작동하는데, 루트란 코드에서 참조되는 전역 변수를 뜻합니다. 모든 루트들은 `acitve` 혹은 `alive`(가비지가 아님)로 표시되며, 이들의 자식들 또한 재귀적(recursively)으로 동일하게 처리됩니다. 즉, 루트에서 접근 가능하다면 가비지가 아니라고 판단됩니다. 2. Sweep: `active`로 표기되지 않은 메모리를 모두 가비지로 간주하여 반환합니다. > [자바스크립트와 V8 엔진의 메모리 관리 프로세스](https://medium.com/naver-place-dev/%EC%9E%90%EB%B0%94%EC%8A%A4%ED%81%AC%EB%A6%BD%ED%8A%B8%EC%99%80-v8-%EC%97%94%EC%A7%84%EC%9D%98-%EB%A9%94%EB%AA%A8%EB%A6%AC-%EA%B4%80%EB%A6%AC-%ED%94%84%EB%A1%9C%EC%84%B8%EC%8A%A4-f45091e696e1)

b7bb682c52bc4be98043c847381fe483

Reference Counting

Tracing garbage collection

null

객체를 참조하는 변수의 수를 추적하는 방법 레퍼런스가 하나 증가/복사 될 때마다 카운트가 1만큼 증가 레퍼런스 카운팅은 매우 심플하고 빠른 가비지 컬렉션 방식이지만, cyclic/circular reference 를 처리하지 못하는 등의 단점 스마트 포인터 cpp

a780282e427c436f8e12ca7f95173926

Stop the world GC

Tracing garbage collection

null

5b16f1ca6ac3450bbf63e3bf56397538

Strong and weak references

Tracing garbage collection

null

7c3454cf4f1843d991bdc68439ec0091

CMS Collector

Mark & Sweep GC

null

## Concurrent Mark & Sweep

5893234d493d4a9a8e14b4a392d44827

Interrupt

Processor Scheduling Notion

May 19, 2021

Alan Jo

Mar 28, 2023

[reverie](https://github.com/facebookexperimental/reverie)

### interception ### Interrupt Notion |Title| |:-:| |[Mode Switching](https://texonom.com/mode-switching-b7f85582f0474555a8c08da922b33610)| |[OS Mode](https://texonom.com/os-mode-b0375e3351064cbd9d220d1c9d44e0e8)| |[Syscall](https://texonom.com/syscall-74f2aa4e9d7141dbad05bae5c88a07e0)| |[Context Switching](https://texonom.com/context-switching-132451f1c9fe4324a1a2bd84e11e7696)| |[Preemptive Scheduling](https://texonom.com/preemptive-scheduling-c4b6e46752f14d69bb769e987ae1c21b)| |[Cooperative Scheduling](https://texonom.com/cooperative-scheduling-93d38df64c4843e68d71e5db4e0f5752)| > [백엔드 개발자들이 알아야할 동시성 4 — 스케줄링](https://choi-geonu.medium.com/백엔드-개발자들이-알아야할-동시성-4-스케줄링-e684331afc77)

d26d34c54d5447a6888ad85089184fb3

Process

Processor Scheduling Notion

Nov 5, 2019

Alan Jo

Seong-lae Cho

May 19, 2023

[Memory Management](https://texonom.com/memory-management-dbf37c509eaf46deb2df8ccd086e386c)

### process - program in execution ### Process Notion |Title| |:-:| |[Process Sync](https://texonom.com/process-sync-45211a317f4f442cbab6b374fdaed630)| |[Process switching ](https://texonom.com/process-switching-7f2486a6184241fba896860610a4d5f1)| |[Process Memory](https://texonom.com/process-memory-53285d27582d49058052939d739e8623)| |[Process Function Call](https://texonom.com/process-function-call-ab41c2af3a5743599fca591ce08b535f)| |[Process State](https://texonom.com/process-state-bbe0f9abe7b240c7a1d2a2d1c3478afd)| |[Process control block](https://texonom.com/process-control-block-f62e2db348e442c580ea6d2084973973)| |[Process descriptor](https://texonom.com/process-descriptor-81614898a94d48058419b80e9e971fae)| |[Process image ](https://texonom.com/process-image-c6384d00d1af41a6bcaa1b238165be17)| |[Process context](https://texonom.com/process-context-e76d70c66cd840b4a247de3310e85416)| |[Process Exit](https://texonom.com/process-exit-2d82b561fba64541bde82c478297a22c)| |[Critical Section](https://texonom.com/critical-section-8469b236528b41efb4088ea3a69a4cc4)| |[Semaphore](https://texonom.com/semaphore-46b11416972249d2a40e6c3e5060304a)| |[Monitor Sync](https://texonom.com/monitor-sync-54b03a3282964bdb88207ec60dfeb1b2)| |[IPC](https://texonom.com/ipc-98010a1077fd45968ce78ae98505b2e5)| |[massage passing](https://texonom.com/massage-passing-ab22575af94b441caad37d2e1697cd59)| |[Mutex](https://texonom.com/mutex-cbfff707d2fe484ea20344f1104b7141)| > memory, device, files, process has each table > **process image **is get by pid > [서버에 걸리는 부하, 추측하지 말고 계측하자](https://injae-kim.github.io/dev/2020/07/09/how-to-check-single-server-load-average.html)

ffd68f9e2b2648759b96047d48d17778

Processor Scheduler

Processor Scheduling Notion

May 19, 2021

Alan Jo

Feb 21, 2023

### Processor Schedulers |Title| |:-:| |[Long-term scheduler](https://texonom.com/long-term-scheduler-5f0783472fa94c478b10b7a1bfbf0192)| |[Short-term scheduler](https://texonom.com/short-term-scheduler-f95f24cbd78d490c8a216dcabb04e4cb)| |[Medium-term scheduler](https://texonom.com/medium-term-scheduler-15b70c90c2f641e696fcc6c90cbd04d9)| ### Processor Scheduler Notion |Title| |:-:| |[Processor Dispatcher](https://texonom.com/processor-dispatcher-dac6ca05f13c4ddc82906bd73f6080df)| ### Processor Queues |Title| |:-:| |[Job Queue ](https://texonom.com/job-queue-92bccac0db114b5c83a0f3ac0d9de17e)| |[Ready Queue](https://texonom.com/ready-queue-c8546adf7d604a73a15b292808bef5b7)| |[Device Queue](https://texonom.com/device-queue-453d40088668423c96d51003b27f9a2e)| - selection function (which) in ready Q - decision mode (when) - end time > - non preem > - preem > [[O/S] CPU Scheduler와 Dispatcher](https://velog.io/@ss-won/OS-CPU-Scheduler%EC%99%80-Dispatcher)

8c124741c1e84d26b0b0a8ad599337f4

Thread

Processor Scheduling Notion

Nov 5, 2019

Alan Jo

Seong-lae Cho

Feb 21, 2023

### Thread Notion |Title| |:-:| |[Thread Safe](https://texonom.com/thread-safe-a6b8ab6ec6e146deafa86b428e5ecfbe)| |[Threads model](https://texonom.com/threads-model-800491618af24626b02f701d43578b66)| |[kernel level thead](https://texonom.com/kernel-level-thead-498f95b4e748492b8e2ed535bf987b0b)| |[Lock convoy](https://texonom.com/lock-convoy-a904af475dcb4bb8a0d99ea8e095de39)| |[Two-Step Dance](https://texonom.com/two-step-dance-b328c6d4bef8483a97e076bbc91b0be8)| |[Priority Inversion](https://texonom.com/priority-inversion-beb6115d6154481db836a5677f877581)| |[Threads Scheduling](https://texonom.com/threads-scheduling-e1315fdaa06740c6ab54f097edeba1d7)| |[Deadlock](https://texonom.com/deadlock-8c118f296cc1477aac5c59c75cec8235)| |[CAS](https://texonom.com/cas-b3541e0665b74cf7aa4f91a7b202fde2)| ### Thread Types |Title| |:-:| |[user level thread](https://texonom.com/user-level-thread-6f68e5d2dff240359b9b494534598b29)| > [좀 생소한 스레드 관련 문제](https://m.blog.naver.com/PostView.nhn?blogId=jjoommnn&logNo=130036504092&proxyReferer=https:%2F%2Fwww.google.com%2F)

d704a229d3b94619963c61bb742b822b

****Context Switching****

Interrupt Notion

Mar 28, 2023

Alan Jo

Mar 28, 2023

### ****Context Switching Notion**** |Title| |:-:|

132451f1c9fe4324a1a2bd84e11e7696

****Cooperative Scheduling****

Interrupt Notion

Mar 28, 2023

Alan Jo

Mar 28, 2023

협조적 스케줄링

93d38df64c4843e68d71e5db4e0f5752

Mode Switching

Interrupt Notion

May 19, 2021

Alan Jo

Feb 21, 2023

b7f85582f0474555a8c08da922b33610

OS Mode

Interrupt Notion

Nov 5, 2019

Alan Jo

Seong-lae Cho

Feb 21, 2023

# hardware protection 1. dual mode - user mode / kernel mode - need hardware support 2. hardware protect protected processor register in intel **CPL(current), RPL(request), DPL(descriptor)** > supervisor = system = monitor = kernel > before mode > intel real mode, protected mode # Mode change ### 1. hardware(external) interrupt - i/O interrupt - device interrupt ← timer intterupt ### 2. software(internal) interrupt - exception → exception handler ### 3. system call # system call → system call handler (by system call interface - inplement of wapper funct in libc.a) ### Method > int 0x80(vector) or sysenter > 1. same reg to KM stack > 2. invoke system call service routine > 3. exit handler > syscall_exit or sysexit > <details><summary> > > example</summary> > 1. fork () > 2. fork{} in libc.a > 3. movl 2, %eax - 2 means fork, eax is register > 4. int 0x80(system call) > 5. go to IDT(interrupt desciptor table) > 6. go to kernel mode function > 7. save all in stack > 8. check eax to function > 9. find sys call table that (2, fork) > 10. do ref_from_sys_call in entry.S > > </details> > > ### kind > > [System call](https://texonom.com/system-call-f36194e943f3498593520460b36c55aa) > - process management system call - file management system call - file system management system call > &system_call - handler address # Bridge there is breidge like north and south to control traffic > no bridge in embedded # VM - hosted - up on host OS - stand alone - up on hardware [dual booting](https://texonom.com/dual-booting-24656ee018014e6388b841b9f579c30c)

b0375e3351064cbd9d220d1c9d44e0e8

Preemptive Scheduling

Interrupt Notion

Mar 28, 2023

Alan Jo

Mar 28, 2023

선점형 스케쥴링

c4b6e46752f14d69bb769e987ae1c21b

Syscall

Interrupt Notion

Feb 21, 2023

Alan Jo

Feb 21, 2023

74f2aa4e9d7141dbad05bae5c88a07e0