Performance Diagnosis with Logs and Metrics

This project introduced two approaches on analyzing performance bugs for system logs and metrics: keyword-based and two-indicator. Keyword-based approach ranks logs based on the weighting of automatically selected keywords. Two-indicator approach utilizes the hotness of timing and the uniqueness of wording of logs to rank them.

Code Structure

• reproduce: This folder contains scripts and results of performance bug reproduces.
  • manual: The experiments with manual triggering of bug and collecting of data.
  • automated: Fully automated bug reproduce with log and metric collection.
• diagnosis: The implementation and experiment result of both approaches.
  • v1: For keyword-based approach.
  • v2: For two-indicator approach.

Example Usage

1. Create Python vitural environment and install dependencies

   $ cd perf-dx
   $ python -m venv .venv
   $ .venv/bin/pip install -r requirements.txt
   

2. To run the two-indicator log analysis, you have to set the environment variable RECORD_ROOT as the absolute path to the folder contains bug reproduce results.

   $ export RECORD_ROOT=/home/myuser/perf-dx/reproduce/HADOOP-11252
   $ .venv/bin/python diagnosis/v2/main.py /tmp/result.csv
   

3. The basic result of analysis will be print on the standard output and detials of uniquness and hotness can be found in /tmp/result.csv in this case.