On Tue, Aug 03, 2010 at 21:58:56 -0600, Zooko O'Whielacronx wrote:
> > Are you using the canned tahoe-lafs repository that one fetches with
> > darcs-benchmark, or are you using a more recent copy.
> 
> The canned one, to make the results more comparable to other people's
> benchmarks.

Hmm, how interesting...

performance regression in record?!

> > I don't know what the best practices are, but it does sound like this is
> > a more conservative and informative way to report results, at least in
> > the graphs.  If nobody complains, I may tweak darcs-benchmark so that
> > its graphs use the mean+sd figure.
> 
> Using mean + sd or mean + 2 * sd is a reasonable approximation of a
> percentile requirement.
> 
> [1] 
> http://s3.amazonaws.com/AllThingsDistributed/sosp/amazon-dynamo-sosp2007.pdf

HEAD darcs-benchmark uses mean + sd everywhere.
In the tables, you'll see a number, followed by the stdev 401 ms (32 ms)
The first number is the mean + sd figure; to get the mean, just subtract
the stddev.

If somebody with better statistical fu says mean + 2 sd is better, it's
an easy switch to flip

> I was thinking maybe a command to run everything at least 20 times so
> that no '?' or '~' would appear (and the results would be more
> statistically reliable).

Maybe.  The reason I avoided that was to make it easy for people to be
able to interrupt darcs-benchmark and still get some good results.

HEAD darcs-benchmark has a --converge flag.  The granularity is only
at the (repo, benchmark) level, which is still better than nothing.
If you run darcs-benchmark run --converge, for each repository, it will
only run benchmarks with less that 20 data points collected.

Perhaps a simple tweak would be to have a special exit code indicating
that there are no more benchmarks to run; then folks can loop.

> >  - fast darcs get over networks - Alexey's Summer of Code project will
> >   be refined over time.  One aim is to reduce the number of files that
> >   darcs has to download.
> 
> Yay! :-)

Alexey's recent blog post (see planet Darcs is quite promising).
I guess it's too late for darcs 2.5, but maybe that's for the better.
It will give us 6 more months (until darcs 2.8) to make sure we have
it right.

-- 
Eric Kow <http://www.nltg.brighton.ac.uk/home/Eric.Kow>
For a faster response, please try +44 (0)1273 64 2905.

Attachment: signature.asc
Description: Digital signature

_______________________________________________
darcs-users mailing list
[email protected]
http://lists.osuosl.org/mailman/listinfo/darcs-users

Reply via email to