Is your feature request related to a problem? Please describe.
We need to have a standard way of measuring performance of the core tools. This is quite helpful to know the impact of adding extra features/analysis that is likely to increase the work needed by the tools
At least, we need:
Throughput: How much time it takes to analyze a set of eventlogs
Memory: total allocated objects, Gc time (although it is not strong indications of poor performance)
Later, we can increase the functionality to include Microbenchmarking for different standalone components.
[ ] Use the Benchmark class defined in org.apache.spark.benchmarks
[ ] Extend the spark benchmark classes to measure GC
[ ] Extend he spark benchmark classes to generate report in structured format (i.e., JSON)
[ ] Define a main method that can be called from the CLI to run the BenchmarkSuite
Is your feature request related to a problem? Please describe.
We need to have a standard way of measuring performance of the core tools. This is quite helpful to know the impact of adding extra features/analysis that is likely to increase the work needed by the tools
At least, we need:
Later, we can increase the functionality to include Microbenchmarking for different standalone components.
org.apache.spark.benchmarks