Virtual Memory B: Objec5ves

Transcription

1 Virtual Memory B: Objec5ves Benefits of a virtual memory system" Demand paging, page-replacement algorithms, and allocation of page frames" The working-set model" Relationship between shared memory and memory-mapped files" To explore how kernel memory is managed" Slides based on Text by Silberschatz, Galvin, Gagne Berkeley Opera7ng Systems group S. Pallikara Other sources Yashwant K Malaiya CS370 Opera7ng Systems Fall

2 Page Replacement Prevent over- alloca7on of memory by modifying page- fault service rou5ne to include page replacement Use modify (dirty) bit to reduce overhead of page transfers only modified pages are wrihen to disk 2

3 Need For Page Replacement Now no place for B 3

4 Basic Page Replacement 1. Find the loca5on of the desired page on disk 2. Find a free frame: - If there is a free frame, use it - If there is no free frame, use a page replacement algorithm to select a vic7m frame - Write vic5m frame to disk if dirty 1. Bring the desired page into the (newly) free frame; update the page and frame tables 2. Con5nue the process by restar5ng the instruc5on that caused the trap Note now poten5ally 2 page transfers for page fault increasing EAT 4

5 5 Page Replacement

6 Page and Frame Replacement Algorithms Frame- alloca7on algorithm determines How many frames to give each process Which frames to replace Page- replacement algorithm Want lowest page- fault rate on both first access and re- access Evaluate algorithm by running it on a par5cular string of memory references (reference string) and compu5ng the number of page faults on that string String is just page numbers, not full addresses Repeated access to the same page does not cause a page fault Results depend on number of frames available In all our examples, the reference string of referenced page numbers is 7,0,1,2,0,3,0,4,2,3,0,3,0,3,2,1,2,0,1,7,0,1 6

7 7 Graph of Page Faults Versus The Number of Frames

8 First- In- First- Out (FIFO) Algorithm Reference string: 7,0,1,2,0,3,0,4,2,3,0,3,0,3,2,1,2,0,1,7,0,1 3 frames (3 pages can be in memory at a 5me per process) 8 15 page faults" Can vary by reference string: consider 1,2,3,4,1,2,5,1,2,3,4,5 Adding more frames (3 to 4) can cause more page (9 to 10) faults! Belady s Anomaly How to track ages of pages? Just use a FIFO queue Belady was here at CSU. Guest in my CS530!

9 FIFO Illustra5ng Belady s Anomaly 16 number of page faults number of frames

10 Op5mal Algorithm Replace page that will not be used for longest period of 5me 9 is op5mal for the example (FIFO was 15) How do you know this? Would have been nice if we could Can t read the future Used for measuring how well your algorithm performs 10

11 Least Recently Used (LRU) Algorithm Use past knowledge rather than future Replace page that has not been used in the most amount of 5me Associate 5me of last use with each page 12 faults beher than FIFO but worse than OPT Generally good algorithm and frequently used But how to implement? 11

12 How we got here.. External fragmenta5on > Paging > Low degree of mul5programming > Demand paging Page Faults > Page replacement algorithm Page buffering Frame alloca5on Working sets 12

13 Ques5ons from last 5me Demand paging: page loaded on demand only Zero- fill: wiped clean for security EAT equa5on: page fault/no page fault Copy- on- write: private copy for wri5ng How does an algorithm know which page it will not need soon. My addi5on Malaiya s anomaly ( anomalous reverse conduc5on 89) 13

14 LRU Algorithm (Cont.) Counter implementa5on Every page table entry has a counter; every 5me page is referenced through this entry, copy the clock into the counter 5me of use field When a page needs to be changed, look at the counters to find smallest value Search through page table needed Stack implementa5on Keep a stack of page numbers in a double link form: Page referenced: move it to the top requires mul5ple pointers to be changed But each update more expensive No search for replacement: bohom page is vic5m LRU and OPT are cases of stack algorithms that don t exhibit Belady s Anomaly 14

15 Use Of A Stack to Record Most Recent Page References reference string a b stack before a stack after b Most recently accessed page moved to the top BoHom page candidate for replacement 15

16 LRU Approxima5on Algorithms LRU needs special hardware and s5ll slow LRU approxima7on: Reference bit With each page associate a bit, ini5ally = 0 When page is referenced bit set to 1 Replace any with reference bit = 0 (if one exists) 1 = recently used We do not know the order, however (improvements possible with mul5ple bits) LRU approxima7on: Second- chance algorithm Generally FIFO, plus hardware- provided reference bit Clock replacement If page to be replaced has Reference bit = 0 - > replace it reference bit = 1 then: set reference bit 0, leave page in memory replace next page, subject to same rules 16

17 Reference bit with shiq register ShiU register Reference bit for page ShiU register auer OS 7mer interrupt

18 Reference bit with shiq register Interpre5ng the reference bits Interpret 8- bit bytes as unsigned integers Page with the lowest number is the LRU page : Not used in last 8 periods : Used 4 5mes in the last 8 periods used more recently than

19 Second- chance algorithm LRU approxima7on: Second- chance algorithm Generally FIFO, plus hardware- provided reference bit avoids throwing out a heavily used page Clock replacement Examine reference bit of next page Reference bit = 0 - > page is old, evict it reference bit = 1 then give it another chance: set reference bit 0, leave page in memory Examine page, subject to same rules 1 = recently used 19

20 Second- Chance (clock) Page- Replacement Algorithm reference bits pages reference bits pages 0 0 next victim nd chance 0 0 Also called Clock algo 1 2 nd chance 0 Hand points to oldest page circular queue of pages (a) circular queue of pages Pointer indicates which page is to be replaced next. When a frame is needed, pointer advances un5l it finds page with a 0 reference bit. As it advances, it clears the reference bits. Once a vic5m page is found, page is replaced, and new page is inserted in circular queue in that posi5on. 20 (b)

21 Enhanced Second- Chance Algorithm Improve algorithm by using reference bit and modify bit (if available) in concert Take ordered pair (reference, modify) 1. (0, 0) neither recently used not modified best page to replace 2. (0, 1) not recently used but modified not quite as good, must write out before replacement 3. (1, 0) recently used but clean probably will be used again soon 4. (1, 1) recently used and modified probably will be used again soon and need to write out before replacement When page replacement called for, use the clock scheme but use the four classes replace page in lowest non- empty class Might need to search circular queue several 5mes 21

22 Coun5ng Algorithms Keep a counter of the number of references that have been made to each page Not common Least Frequently Used (LFU) Algorithm: replaces page with smallest count Most Frequently Used (MFU) Algorithm: based on the argument that the page with the smallest count was probably just brought in and has yet to be used 22

23 Page- Buffering Algorithms Buffering: Keep a pool of free frames Then frame available when needed, not found at fault 5me Read page into free frame and select vic5m to evict and add to free pool When convenient, evict vic5m Pro- ac5ve: Keep list of modified pages When backing store otherwise idle, write pages there and set to non- dirty If a page is selected for replacement increase likelihood of that page being clean Reuse, keep free frame contents intact and note what page they held If same page referenced again before reused, no need to load contents again from disk 23

24 Ques5ons from last 5me LRU: Would a queue be a good data structure? Page Buffering Algorithms : ideas to reduce demand paging overheads Why number of page faults is an important number? Buffering: Keep a pool of free frames Bring in frame now, evict vic5m later Pro- ac5ve: Save modified pages when convenient Don t wait for a page fault Keep freed frame contents and note what page they held In case same page referenced again before reuse 24

25 Applica5ons and Page Replacement All of these algorithms have OS guessing about future page access Some applica5ons have beher knowledge of future pages i.e. databases Memory intensive applica5ons can cause double buffering OS keeps copy of page in memory as I/O buffer Applica5on keeps page in memory for its own work 25

26 Alloca5on of Frames How do you divvy up free memory among processes? Each process needs minimum number of frames Example: IBM pages to handle SS MOVE instruc5on: instruc5on is 6 bytes, might span 2 pages 2 pages to handle from 2 pages to handle to Maximum of course is total frames in the system Two major alloca5on schemes fixed alloca5on priority alloca5on Many varia5ons in alloca5on approaches 26

27 A. Fixed Alloca5on Equal alloca7on For example, if there are 100 frames (aqer alloca5ng frames for the OS) and 5 processes, give each process 20 frames Have to keep some as free frame buffer pool Propor7onal alloca7on Allocate according to the size of process Dynamic as degree of mul5programming, process sizes change s S = s a i i = size of process i m = total number of = allocation for p i p i frames si = m S m = 64 s1 =10 s 2 =127 a 1 = a 2 =

28 Priority Alloca5on Use a propor5onal alloca5on scheme using priori5es rather than size If process P i generates a page fault, select for replacement one of its frames: Local grab for replacement a frame from a process with lower priority number: Global 28

29 Global vs. Local Alloca5on Global replacement process selects a replacement frame from the set of all frames; one process can take a frame from another But then process execu5on 5me can vary greatly But greater throughput so more common Local replacement each process selects from only its own set of allocated frames More consistent per- process performance But possibly underu5lized memory 29

30 Non- Uniform Memory Access NUMA So far all memory accessed equally Many systems are NUMA speed of access to memory varies Consider system boards containing CPUs and memory, interconnected over a system bus Op5mal performance comes from alloca5ng memory close to the CPU on which the thread is scheduled And modifying the scheduler to schedule the thread on the same system board when possible Idea used by Solaris 30

31 Problem: Thrashing If a process does not have enough pages, the page- fault rate is very high Page fault to get page Replace exis5ng frame But quickly need replaced frame back This leads to: Low CPU u5liza5on, leading to Opera5ng system thinking that it needs to increase the degree of mul5programming leading to Another process added to the system Thrashing a process is busy swapping pages in and out 31

32 32 Thrashing (Cont.)

33 Demand Paging and Thrashing Why does demand paging work? Locality model Process migrates from one locality to another Locali5es may overlap Why does thrashing occur? Σ size of locality > total memory size Limit effects by using local or priority page replacement Working sets: iden5fy locality 33

34 Locality In A Memory- Reference PaHern page numbers memory address execution time

35 Working- Set Model Δ working- set window a fixed number of page references Example: 10,000 instruc5ons WSS i (working set of Process P i ) = total number of pages referenced in the most recent Δ (varies in 5me) if Δ too small will not encompass en5re locality if Δ too large will encompass several locali5es if Δ = will encompass en5re program D = Σ WSS i total demand frames Approxima5on of locality if D > m Thrashing Policy if D > m, then suspend or swap out one of the processes Δ = 10 35

36 Keeping Track of the Working Set Approximate with interval 5mer + a reference bit Example: Δ = 10,000 units Timer interrupts aqer every me units Keep in memory 2 bits for each page Whenever a 5mer interrupts copy and sets the values of all reference bits to 0 If one of the bits in memory = 1 page in working set Why is this not completely accurate? Improvement = 10 bits and interrupt every me units 36

37 Page- Fault Frequency More direct approach than WSS Establish acceptable page- fault frequency (PFF) rate and use local replacement policy If actual rate too low, process loses frame If actual rate too high, process gains frame page-fault rate increase number of frames upper bound lower bound decrease number of frames number of frames 37

38 Working Sets and Page Fault Rates Direct rela5onship between working set of a process and its page- fault rate Working set changes over 5me Peaks (working set changed) and valleys (working set stable) over 5me Transi5ons: 3 working sets 38

39 Memory- Mapped Disk Files Memory- mapped file I/O allows file I/O to be treated as rou5ne memory access by mapping a disk block to a page in memory A file is ini5ally read using demand paging A page- sized por5on of the file is read from the file system into a physical page Subsequent reads/writes to/from the file are treated as ordinary memory accesses Simplifies and speeds file access by driving file I/O through memory rather than read() and write() system calls Also allows several processes to map the same file allowing the pages in memory to be shared But when does wrihen data make it to disk? Periodically and / or at file close() 5me For example, when the pager scans for dirty pages 39

40 Ques5ons from last 5me Frame alloca5on for a process Threshing: when a process has too few frames Working set update: to avoid threshing, make sure working set is in memory Which is the best choice for page replacement algorithm? Transfers between memory and disk: DMA Memory mapped files: delays 40

41 Memory- Mapped File Technique for all I/O Some OSes uses memory mapped files for standard I/O Process can explicitly request memory mapping a file via mmap() system call Now file mapped into process address space For standard I/O (open(), read(), write(), close()), mmap anyway But map file into kernel address space Process s5ll does read() and write() Copies data to and from kernel space and user space Uses efficient memory management subsystem Avoids needing separate subsystem COW can be used for read/write non- shared pages Memory mapped files can be used for shared memory (although again via separate system calls) 41

42 Memory Mapped Files process A virtual memory process B virtual memory 4 2 physical memory disk file Disk File uses 6 blocks Page tables used for mapping 42

43 Shared Memory via Memory- Mapped I/O process 1 process 2 shared memory memory-mapped file shared memory shared memory 43

44 Alloca5ng Kernel Memory Treated differently from user memory Oqen allocated from a free- memory pool Kernel requests memory for structures of varying sizes Process descriptors, semaphores, file objects etc. Oqen much smaller than page size Some kernel memory needs to be con5guous I.e. for device I/O Two approaches (skipped) Buddy system Slab allocator 44

45 Other Considera5ons - - Prepaging Prepaging To reduce the large number of page faults that occurs at process startup Prepage all or some of the pages a process will need, before they are referenced But if prepaged pages are unused, I/O and memory was wasted Assume s pages are prepaged and α of the pages is used Is cost of s * α save pages faults > or < than the cost of prepaging s * (1- α) unnecessary pages? α near zero prepaging loses 45

46 Other Issues Page Size Some5mes OS designers have a choice Especially if running on custom- built CPU Page size selec5on must take into considera5on: Fragmenta5on: smaller is beher Page table size: larger is beher I/O overhead: large beher Locality: smaller is beher Always power of 2, usually in the range 2 12 (4,096 bytes) to 2 22 (4,194,304 bytes) On average, growing over 5me 46

47 Other Issues TLB Reach TLB Reach - The amount of memory accessible from the TLB TLB Reach = (TLB Size) X (Page Size) Ideally, the working set of each process is stored in the TLB Otherwise there is a high degree of page faults Increase the Page Size to increase reach? This may lead to an increase in fragmenta5on as not all applica5ons require a large page size Provide Mul5ple Page Sizes This allows applica5ons that require larger page sizes the opportunity to use them without an increase in fragmenta5on 47

48 Other Issues Program Structure Program structure int[128,128] data; i: row, j: column Each row is stored in one page Program 1 for (j = 0; j <128; j++) for (i = 0; i < 128; i++) data[i,j] = 0; 128 x 128 = 16,384 page faults Program 2 for (i = 0; i < 128; i++) for (j = 0; j < 128; j++) data[i,j] = 0; 128 page faults 48

49 Other Issues I/O interlock I/O Interlock Pages must some5mes be locked into memory Consider I/O - Pages that are used for copying a file from a device must be locked from being selected for evic5on by a page replacement algorithm Pinning of pages to lock into memory 49

50 Opera5ng System Examples (Skipped) Windows Solaris 50