bbengfort / compile-workload

Python script to create a compiler workload and evaluate external performance.

Compile Workload

Python script to create a compiler workload and evaluate external performance.

This script is intended to be used with MemFS to create a workload for evaluating system performance. The primary workload script does the following activities in the specified directory:

1. Clone a repository
2. Make/build in the repository
3. Stat every file
4. Report time and stats

Because compilation of a project is different on every system, project-specific steps are embedded in the workload file. The following projects are supported:

1. Redis (76MB, 591 files, 137,475 LOC)
2. Postgres (422MB, 4,807 files, 910,948 LOC)
3. Nginx (62MB, 440 files, 155,056 LOC)
4. Apache Web Server (362MB, 4,059 files, 503,006 LOC)
5. Ruby (197MB, 3,281 files, 918,052 LOC)
6. Python 3 (382MB, 3,570 files, 931,814 LOC)

Running a workload is as follows:

$ python3 workload.py -o results.csv -p nginx /tmp/testdir

This will run the workload on the Nginx repository, building in /tmp/testdir and appending the results to results.csv. Various options and defaults are provided and can be further inspected with:

$ python3 workload.py --help

Note that this script depends on many system and compiling dependencies to be available. Because they were already available on my system, I don't necessarily have a list of that I can expose through a requirements file. However, at a minimum, Git is required as are Xcode developer tools on a Macbook Pro.

For testing MemFS, I've created a simple script that runs through a single instance of the testing protocol. This should be used with care, however, as it is built for a specific system.