CRISP: Critical Path Analysis of Microservice Traces

This repo contains code to compute and present critical path summary from Jaeger microservice traces. To use first collect the microservice traces of a specific endpoint in a directory (say traces). Let the traces be for OP operation and SVC service (these are Jaeger termonologies). python3 process.py --operationName OP --serviceName SVC -t <path to trace> -o . --parallelism 8 will produce the critical path summary using 8 concurrent processes. The summary will be output in the current directory as an HTML file with a heatmap, flamegraph, and summary text in criticalPaths.html. It will also produce three flamegraphs flame-graph-*.svg for three different percentile values.

The script accepts the following options:

python3 process.py --help
usage: process.py [-h] -a OPERATIONNAME -s SERVICENAME [-t TRACEDIR] [--file FILE] -o OUTPUTDIR
                  [--parallelism PARALLELISM] [--topN TOPN] [--numTrace NUMTRACE] [--numOperation NUMOPERATION]

optional arguments:
  -h, --help            show this help message and exit
  -a OPERATIONNAME, --operationName OPERATIONNAME
                        operation name
  -s SERVICENAME, --serviceName SERVICENAME
                        name of the service
  -t TRACEDIR, --traceDir TRACEDIR
                        path of the trace directory (mutually exclusive with --file)
  --file FILE           input path of the trace file (mutually exclusivbe with --traceDir)
  -o OUTPUTDIR, --outputDir OUTPUTDIR
                        directory where output will be produced
  --parallelism PARALLELISM
                        number of concurrent python processes.
  --topN TOPN           number of services to show in the summary
  --numTrace NUMTRACE   number of traces to show in the heatmap
  --numOperation NUMOPERATION
                        number of operations to show in the heatmap

Dataset

• We released the artifact of the original CRISP paper at https://zenodo.org/records/13956078.
• We released 1.5 Million production traces along with our paper The Tale of Errors in Microservices at https://zenodo.org/records/13947828.

Please cite our paper if you use the dataset in your research.