- Cache Simulation
- Date 12/19/2024
- Alex Taylor
In this program, I designed a cache implementation using the linked list data structure. The program can simulate the behavior of a single-level cache and a two-level cache. I also created a test class to verify the functionality of the cache. This was an assignment for CS321, my data structures and algorithms class. All code was written by me.
- Cache.java - A collection of methods used to simulate cache behavior.
- Test.java - Driver class used to test the correctness of cache implementation.
- README - A brief overview of the program, how to run and compile it, related concepts, and testing information.
In a terminal window, navigate to the directory containing all source files, compile the Test class with the command:
$ javac Test.java
To test a single cache implementation, run the command:
$ java Test 1 <1st-level cache size> <"textfile name">
To test a multi-level(two) cache implementation, run the command:
$ java Test 2 <1st-level cache size> <2nd-level cache size> <"textfile name">
The console output will give the results after the program finishes.
This program implements a cache simulation using a linked list data structure. A cache is a storage in memory, and data objects in the cache are frequently accessed by the computer. This cache program is a simulation of that behavior. The cache uses an MRU scheme. MRU stands for "Most Recently Used", where if a user needs a data item from the cache, the item is returned to the user and in turn placed at the top of the cache. If the cache is full, the oldest entry in the cache is removed.
This program supports the implementation of two caches. The 2nd level cache is usually bigger than the 1st level cache. The 2nd level cache contains all data that is in the 1st level cache. This is called the multilevel inclusion property.
The global hit ratio for a single-level cache is defined by HR = (# of cache 1 hits / # of cache 1 references)
The global hit ratio for a multi-level cache is defined by HR = (# of cache 1 hits + # of cache 2 hits) / (# of cache 1 references)
The reason the global hit ratio for a multi-level cache is divided by only the number of references to the first cache is because we have to consider the cache as a black box. This means the (# of cache 1 references = # of global references).
Test.java is the driver class and in turn the test class for the Cache class. I was able to verify the cache's functionality by comparing the outputs of my cache's hit and reference ratios output by Test.java to the values given to me by my professor.
Program runtime is reported after each cache simulation is run.
The runtime for:
"java Test 2 1000 2000 Encyclopedia.txt" --> ≈ 7676 milliseconds
"java Test 2 1000 5000 Encyclopedia.txt" --> ≈ 15092 milliseconds
"java Test 2 <> <> small.txt" --> ≈ 1 millisecond
For reference, "small.txt" is a 2 line file, whereas "Encyclopedia.txt" is ≈ 147,000 lines long.