[squid-users] Squid test-suite / benchmarks

Alex Rousskov rousskov at measurement-factory.com
Thu Jun 7 17:20:38 UTC 2018

On 06/07/2018 04:17 AM, Stoica Bogdan Alexandru wrote:

> We’re a small research team interested in benchmarking Squid for a
> research project.

> Ideally, we would like to have good code coverage while doing so.

> Are there any good benchmarks used for such purpose?

Performance benchmarks usually focus on things other than code coverage.
It is very difficult to write a quality benchmark for a proxy, even
without code coverage as a goal!

One the other hand, a decent proxy benchmark has enough knobs to tickle
most "interesting" code paths in Squid (or any other proxy). Web
Polygraph[1] (mentioned on this thread earlier) is a good example -- you
can trigger cache revalidation, simulate heavy tailed hit distributions
that stress disk caching, exercise the code that handles aborted
transactions, persistent connection races, etc., etc.

> Or, even better, is
> there a more comprehensive test suite apart from the one Squid comes with?

Squid does not come with a comprehensive test suite (yet) and the tests
distributed with Squid are not performance tests (a.k.a. "benchmarks").
If you are looking for functionality rather than performance testing,
then there is Co-Advisor[2]. Squid is tested with Co-Advisor, but those
tests have not been automated (yet).

  [1] http://www.web-polygraph.org/
  [2] http://coad.measurement-factory.com/


P.S. Disclaimer: The company I work for is responsible for both of the
test tools mentioned above.

More information about the squid-users mailing list