Download Memory Allocation: Techniques and Algorithms - Prof. Indranil Gupta and more Study notes Computer Science in PDF only on Docsity!
1
Memory Allocation
Indranil Gupta (Indy)
2
Reading: Chapter 22
3
Demand Paging - Reminder
Load M
Free frame
Page table
VM
ref
Page fault
Reference V R D W ReE Frame Address
Re fe re nce VRDW ReE Frame Add re ss
4
LRU Approximation Algorithms
Additional Reference Bits Algorithm
Keep n bits for each page in a table in memory
Each reference sets highest order bit
Periodically, shift bits right dropping the lowest
bit)
Use value as 8 bit unsigned integer
Victim Selection: Page with the lowest value of
reference counter
The value may not be unique, so use FIFO to
resolve conflicts
LRU Approximation Algorithms
Second Chance
FIFO (clock) replacement algorithm
Uses reference bit in each page table entry
Initially 0, set to 1 on a reference
Pages are kept in FIFO order using a circular list
If page is selected as prospective victim, inspect
the reference bit
If reference bit = 0, replace the page If page to be replaced (in clock order) has reference bit = 1, then Set reference bit to 0 Leave page in memory Replace next page (in clock order) subject to same rules
Allocation of Page Frames
Scenario
Several physical pages allocated to processes
A, B, and C. Process B page faults.
Which page should be replaced?
Allocating memory across processes?
Does every process get the same fraction of
memory?
Different fractions?
Should we completely swap some processes out
of memory?
7
Allocation of Page Frames
Each process needs a minimum number of
pages to function at all
Want to make sure that all processes that are
loaded into memory can make forward progress
Example: IBM 370 โ 6 pages to handle SS
MOVE instruction:
instruction is 6 bytes, might span 2 pages 2 pages to handle from 2 pages to handle to
8
Fixed Allocation
Allocate a minimum but fixed number of
frames per process
Consider minimum requirements, e.g. on
previous slide
Include an extra page for paging out and one for
paging in
9
Equal Allocation
Allocate an equal number of frames per job
Example
100 frames 5 processes Each process gets 20 frames
Concerns
But jobs use memory unequally
High priority jobs have same number of page
frames and low priority jobs
Degree of multiprogramming might vary
10
Proportional Allocation
Allocate a number of frames per job
proportional to โjob sizeโ
How do you determine job size
Run command parameters? Dynamically?
Priority Allocation
May want to give high priority process more
memory than low priority process.
Use a proportional allocation scheme using
priorities instead of size
Disadvantage?
Allocation of Page Frames
Possible Replacement Approaches
Local replacement
Global replacement
Local Replacement: Per
Process
Each process has separate pool of pages Fixed number of pages (e.g., Digital VMS) Fixed fraction of physical memory (1/P) Proportional to size of allocated address space Page fault in one process only replaces pages of that process Perform replacement (e.g., LRU) over only those pages Advantage No interference across processes Disadvantage Potentially inefficient allocation of memory How to handle sharing of pages across processes?
19
Thrashing
Studentโs analogy to thrashing: Too many courses
Drop a course
OS solution: Admission control
Determine how much memory each process needs Long-term scheduling policy Run only those processes whose memory requirements can be satisfied
20
Working Set
Question
How much memory does a process need to
keep the most recent computation in memory
with very few page faults?
How can we determine this?
Determine the working set of a process
The principle of locality
A program clusters its access to data and text temporally A recently accessed page is more likely to be accessed again E.g., sequential ops - reading arrays, code pages
21
Working Set (1968, Denning)
Need
Set of pages process needs to avoid thrashing Requires knowing the future
Working set: realistic implementation
Pages referenced by process in last ฮค seconds of execution Approximates locality principle
22
Working Set (1968, Denning)
Need
Set of pages process needs to avoid thrashing Requires knowing the future
Working set
Pages referenced by process in last ฮค seconds of execution Approximates locality
Using working set sizes
Cache partitioning: give each app enough space for WS Page replacement: preferentially discard non-WS pages Scheduling: process not executed (ready or running) unless WS in memory
Working Set Size
At least allocate this many frames for this process
Working Set Size
Choosing T
T too small
Will not encompass entire locality
T too large
Will encompass several localities
T = โ
Will encompass entire program
25
Working sets of real programs
Typical programs have phases
Working set size
transition
stable
transition
stable
transition
stable
26
Working Set in Action to
Prevent Thrashing
Algorithm
If number free page frames > working set
of some suspended processi
Activate processi and map in its working set
If working set size of some processk
increases and no page frame is free
suspend processk and release all its pages
27
Working Set Implementation
Tracking the working set
Moving window over reference string Approximate with Interval timer + a reference bit
Example: T = 10,
Keep in memory 2 bits for each page Indicates if page was used within last 10,000 references Timer interrupts after every 5000 time units Shift 0 right If one of the bits in memory = 1 Page is in working set
Not completely accurate - cannot tell where
reference occurred.
Improvement - 10 bits and interrupt every 1000 time units (^28)
Page Fault Frequency
Working Set
Approximation of pure working set
Assumption
If the working set is correct Not many page faults
Approach
Control thrashing by establishing acceptable page-fault rate
Page Fault Frequency
Working Set
Algorithm
If page fault rate increases beyond assumed knee of curve Increase number of page frames available to process If page fault rate decreases below foot of knee of curve Decrease number of page frames available to process.
Page Size Considerations
Small pages
Large page tables Minimizes internal fragmentation Good for locality of reference (~256) Page tables are larger Disk-seek time dominates transfer time (It takes almost the same time to read a large page as a small page)
Large pages
Significant amounts of a page may not be referenced Enables more data per seek
Real systems (can be
reconfigured)
Windows: default 8 KB Linux: default 4 KB
