Description
Introduction
The aim of this assignment is to improve your learning experience in the page replacement
algorithms. Following our discussion of paging in lectures, this practical will allow you to explore
how real applications respond to a variety of page replacement schemes. Since modifying a real
operating system to use different page replacement algorithms can be quite a technical exercise,
your task will be to implement a program that simulates the behaviour of a memory system
using a variety of paging schemes.
Memory Traces
We provide you with some memory traces to assist you developing your simulator.
Each trace is a series of lines, each listing a hexadecimal memory address preceded by R or
W to indicate a read or a write. There are also lines throughout the trace starting with a #
followed by a process ID and command. For example, a trace file for gcc might start like this:
# gcc
R 0041f7a0
R 13f5e2c0
R 05e78900
R 004758a0
W 31348900
The lines in the trace file beginning with #.
Simulator Requirements
Your task is to write a simulator that reads a memory trace and simulates the action of a virtual
memory system with a single level page table.
Your memory system should keep track of which pages are loaded into memory.
• As it processes each memory event from the trace, it should check to see if the corresponding page is loaded.
• If not, it should choose a victim page in memory to replace.
• If the victim page to be replaced is dirty, it must be saved to disk before replacement.
• Finally, the new page is loaded into memory from disk, and the page table is updated.
This is just a simulation of the page table, so you do not actually need to read and write
data from disk. When a simulated disk read or write occurs, simply increment a counter to keep
track of disk reads and writes, respectively.
You must implement the following page replacement algorithms:
• FIFO: Replace the page that has been resident in memory longest.
• LRU: Replace the page that has been resident in memory oldest.
• ARB: Use multiple reference bits to approximate LRU. Your implementation should use
an a-bit shift register and the given regular interval b. See textbook section 9.4.5.1.
– If two page’s ARB are equal, you should use FIFO (longest) to replace the frame.
• WSARB-1: Combine the ARB page replacement algorithm with the working set model
that keeps track of the page reference frequencies over a given window size (page references)
for each process (see textbook section 9.6.2). It works as follows:
– Associate each page with a reference bits as the shift register (R) to track the reference
pattern over the recent time intervals (for a given time interval length), and an integer
counter (C) of 8 bits to record the reference frequency of pages (in the memory frames)
in the current working set window. R and C are initialized to 0.
– Page replacement is done by first selecting the victim page of the smallest reference
frequency (the smallest value of C), and then selecting the page that has the smallest
reference pattern value in the ARB shift register (the smallest value of R) if there are
multiple pages with the smallest reference frequency.
• WSARB-2: Same as WSARB-1 in combining ARB with the working set model, but selecting a victim for page replacement is done in the reverse order:
– First select the page having the smallest reference patten value in R, and then that
of the smallest reference frequency value in C if there are multiple pages with the
smallest reference patten value.
Note that for WSARB-1 and WSARB-2, same as ARB, if there are multiple pages with the
same smallest values of R and C, victim will be chosen according to FIFO among them.
Test
Arguments
Your code will be compiled using following order in your SVN folder.
g++ -std=c++11 PageReplacement.cpp -o PageReplacement
The simulator should accept arguments as follows:
1. The filename of the trace file
2. The page/frame size in bytes (we recommend you use 4096 bytes when testing).
3. The number of page frames in the simulated memory.
The trace provided should be opened and read as a file, **not** parsed as text input from
stdin.
For example, your code might be run like this:
“ ./PageReplacement input.txt 4096 32 FIFO ” Where:
• The program being run is ‘./PageReplacement’,
• The name of the input file is ‘input.txt’,
• A page is ‘4096’ bytes,
• There are ‘32’ frames in physical memory,
• The page replacement algorithm to use: FIFO / LRU / ARB / WSARB-1 / WSARB-2
If the page replacement algorithm is ARB , it should accept the following additional arguments:
The number of reference bits a (1 ≤ a ≤ 8) used in the shift register (R).
The integer value of regular interval b (1 ≤ b ≤ 12) for register shifting.
If the page replacement algorithm is WSARB-1 or WSARB-2, it should accept the following
additional argument:
The number of a, the shift register (R) uses an a-bit shift register (1 ≤ a ≤ 8).
The integer value of regular interval b (1 ≤ b ≤ 12) for register shifting.
The size of the working set window δ, in page references (b ≤ δ ≤ 256).
For example, your code might be run like this:
“
./PageReplacement input.txt 1024 16 ARB 3 3
”
“
./PageReplacement input.txt 4096 32 WSARB-1 8 3 11
”
Input
We will provide you with a selection of memory traces to assist you developing your simulator.
These will be a mix of specific test cases and real traces from running systems. Each trace is a
series of lines, containing two(2) values that represent memory accesses:
1. A character ‘R’ or ‘W’ that represents whether the memory access is a Read or Write
respectively.
2. A hexadecimal memory address.
A trace may also contain comment lines, # followed by a process Name.
An example of a trace:
# chrome
R 0041f7a0
R 13f5e2c0
R 05e78900
R 004758a0
W 31348900
Output
The simulator should run silently with no output until the very end, at which point it prints
out a summary like this:
events in trace: 1025
total disk reads: 151
total disk writes: 92
page faults: 151
Where:
• “events in trace” is the number of memory accesses in the trace. Should be equal to
number of lines in the trace file that start with R or W. Lines starting with # do not
count.
• “total disk reads” is the number of times pages have to be read from disk.
• “total disk writes” is the number of times pages have to be written back to disk.
• “page faults” is the number of disk reads in a demand paging system. It may be same
with total disk reads in this problem.
We will provide a set of expected outputs(on web-submission) to match the given memory
traces.
Web-submission instructions
• First, type the following command, all on one line (replacing xxxxxxx with your student
ID):
svn mkdir –parents -m “OS”
https://version-control.adelaide.edu.au/svn/axxxxxxx/2020/s2/os/assignment2
• Then, check out this directory and add your files:
svn co https://version-control.adelaide.edu.au/svn/axxxxxxx/2020/s2/os/assignment2
cd assignment2
svn add PageReplacement.cpp
· · ·
svn commit -m “assignment2 solution”
• Next, go to the web submission system at:
https://cs.adelaide.edu.au/services/websubmission/
Navigate to 2020, Semester 2, Operating Systems, Assignment 2. Then, click Tab “Make
Submission” for this assignment and indicate that you agree to the declaration. The
automark script will then check whether your code compiles. You can make as many
resubmissions as you like. If your final solution does not compile you won’t get any marks
for this solution.
• We will test your codes by the following Linux commands:
g++ -std=c++11 PageReplacement.cpp -o PageReplacement
./PageReplacement input.txt 4096 32 FIFO >output.txt
FAQ
(https://myuni.adelaide.edu.au/courses/54560/discussion topics/453115)
1. How are the page numbers and memory addresses connected? Is there a relation to page
size?
This comes down to the fundamentals of paging. Be sure to read and understand section
8.5.1 of the text book before going any further. It will save you much confusion later on.
2. I’m confused by the ARB algorithm. How is it actually supposed to work?
For those of you Googling along at home you may have come across a variety of somewhat
conflicting descriptions of this algorithm. For reference, we’re strictly following this one
outlined in the book:
Figure 1: ARB
And we use a-bit shift register (R), instead of 8-bit shift register in book.
• That trace was run on a 64-bit system which uses a larger address space. If you’re
converting these values to numbers, you may need to use the long long data type.
• In a system running this algorithm, there should never be a case where the interval/quantum for shifting the bits is greater than the number of frames. That behavior
is undefined and we do not expect you to code for it.
• I’m confused by the working set model. How is it actually supposed to work?
Please refer to 9.6.2 of pages 427-429 of the text.
For Working-Set ARB (both WSARB-1 and WSARB-2). Some extra notes:
• We are concerned only with single-process memory allocation, so each input reference
string is only for the same process.
• Ensure you keep track of the process’ working set over the given time window on the
input reference string.
• All page faults should be recorded.
• Assume that the total number of frames allocated to the processes is greater or equal
to the size of its working set window.
• The shift register (R) and reference frequency counter (C) only count the pages which
are in memory frames.
For example:
Time: 1 2 3 4 5 6 7 8 9 10…
request pages: 2 1 3 4 5 2 4 6 1 2…
frames size=4;
window size=4;
At end of time 4, four frames are in a table in memory: 2 1 3 4.
Time: 1 2 3 4 5 6 7 8 9 10…
request pages: 2 1 3 4 5 2 4 6 1 2…
At time 5, a new page 5 need into the frames. And the current working set window
is 1 3 4 5. The 8-bit reference frequency counter (C) will find that the frequency of
page 2 is zero (the smallest value of C) in the current working set, and then the new
page 5 will replace page 2 in the memory table.
