Uncovering Steady State Executions in Java Microbenchmarking with Callgraph Analysis

JMH

The data used in this repo was collected using JMH on the java file JMH/test/src/main/java/org/sample/MyBenchmark.java. The java file runs a recursive binary search to find an element in an array. The JMH was run on 10 forks with 1000 iterations and 0 warm-up iterations. We used the perf to collect the profiler events to create the flame graphs.

---

Run

Library Versions used

• numpy==1.22.0
• karateclub==1.3.3
• networkx==2.6.3
• rpy2==3.0.0
• kneed==0.8.2
• sklearn==1.0.2

To visualize the results from our classifications model please run the classifications.py file. The classifications.py file uses the files graph embeddings the vectors.txt file and the labels from the labels.txt file.

To create the graph embeddings run the file create_graph_vectors.py'. The create_graph_vectors.py' file uses the Folder Folded which contains all the Flame Graphs in a txt format.

To create the graph labels run the file create_graph_labels.py. The create_graph_labels.py file uses the iterDurations.txt to create the labels.txt file. The source code used in this file was retrived from https://github.com/SEALABQualityGroup/steady-state.git.