fcrepo4-labs/fcrepo_perf_analysis

Fcrepo Performance Analysis

Fcrepo Performance

Fcrepo Jmeter Analysis

A collection of R scripts which use jmeter log output generated by the standard commmunity tests. These scripts perform analysis to answer specific questions about fedora-commons repository performance, and produce figures to illustrate the answers.

Requirements

• R
• R Packages: ggplot2, knitr, rmarkdown, svglite
• Vendor Packages:
  • Cairo (libcairo2-dev on debian, cairo on OSX)
  • Pandoc (pandoc on debian)

Use

1. Get your perf.log from running a fcrepo test or use a copy of the test data run by virginia tech as a demo.

   # Pre-existing data from test-4 that can be used as an example
   mkdir raw-data
   wget https://s3.amazonaws.com/f4-performance-results/test4.pu.postgres.tar.gz -O ./raw-data/test.tgz
   tar -C raw-data -xzf raw-data/test.tgz
   

2. Run Analysis. Substitutue example/test4-perf.log with the perf.log from your own tests or the one obtained from S3.

   # Make directories.
   mkdir -p processed-data
   mkdir -p build
   mkdir -p reports
   
   # Run install script to make sure required packages are installed.
   Rscript 00-install-required-packages.r
   
   # Check that input data has expected headers
   #   The result should be the printed statement: `
   #   Header check:
   #   Expected: 12 encountered: 12
   Rscript 01-check-jmeter-log.r example/test4-perf.log
   
   # Run preprocessing and save result into processed-data directory.
   Rscript 10-jmeter-create-objects-preprocess.r example/test4-perf.log > processed-data/subset.csv
   
   # Run the analysis 
   Rscript 20-jmeter-create-objects-analysis.r processed-data/subset.csv
   
   # Produce figures
   Rscript 21-jmeter-create-objects-plots.r processed-data/subset.csv
   
   # Produce report
   Rscript 30-knit-report.r ./build report.md
   

Reports

Generated reports for test runs are in dist.