[Design] [cosmo] Performance test for 0.6.1
jared at wordzoo.com
Mon Apr 9 13:17:23 PDT 2007
>> maybe not ui performance testing, but sync performance is arguably the
>> #1 goal of 0.6.1, so i hope that's addressed in a test plan.
> At the moment the test plan is mostly
> "Is Chandler wall clock time shorter?".
I think this is a pretty reasonable guideline.
Couple of notes:
* It'd be really handy if Chandler Desktop reported, in the log file,
the total sync time (if it doesn't already). For both DAV and MC
styles. A stopwatch will probably work too, if people can follow the
status bar indicators.
* For command-line, testing performance of a (non-changing) DAV-based
sync is really just timing the PROPFIND operation, which we've
well-documented for testing via curl. For MC, we could also do the
MC-get via curl. This would be a simple curl-to-curl comparison.
* But I think we're currently missing a way to publish a monolithic ICS
via MC. There's bcm's command-line MC client, but to publish, one needs
a an EIMML-formatted document (I think). I think we're missing any code
(except Desktop) currently to convert an ICS to EIMML. This would be
helpful to have, in any form.
I don't think it's critical to have high-concurrency testing of MC by
0.6.1 release. Raw performance (of single-threaded ops) would be great,
but seems to be available via Ted's "wall clock in Chandler" or via some
pretty-simple command-line ops.
More information about the Design