Cache Performance Insight

Cache Performance Insight is a framework designed to evaluate the performance of different cache policies with different workloads and configurations.

Build

mkdir build && cd build
cmake .. && make

Usage

./src/main <cache_policy> <buffer_size> <trace_file> <params>?

Example

./src/main -h
usage: ./src/main <cache_policy> <buffer_size> <trace_file> <param_0>
       <cache_policy> -- cache policy in {LRU, LFU, ARC, ARC_2, ARC_3, OPT}
       <buffer_size>  -- buffer size
       <trace_file>   -- path of trace_file

./src/main LRU 65536 ../traces/P1.lis
LRU_CACHE_MANAGER:  buffer_size:65536 hit_count:11011495 miss_count:21043978 hit_rate:34.3514%

Unittest & Coverage

ctest && make coverage

Scripts & Visualization

Install

pip install -r requirement

Run

./scripts/test_runner.py

Find plot in local/plot.png:

TODO

Cache policies

• LRU
• LFU
• ARC
• OPT
• ARC-2
• ARC-3
• MRF
• hybrid LRU-LFU
• ...

Features

• cmake build
• unittest & code coverage
• test scripts
• result visualization
• log
• more parameters (e.g. scales, policies, unique keys)
• ...

Metrics

• cache hit rate
• runtime
• actual memory allocation
• concurrency
• real time cache distribution
• ...