On Tuesday 20 March 2007 15:26, John Peacock wrote:
> Malcolm Rowe wrote:
> > My results:
> > - With one exception, sharding is always more expensive than not
> > sharding (because it adds an extra directory lookup). The difference
> > is negligible, however, because we're still talking about microseconds
> > per lookup.
> Thank you for taking the time to benchmark something, rather than
> relying on Magic Numbers (something I'm severely allergic to). ;-)
Yes, thank you very much.
Regarding your numbers - I think that linux might be a bad candidate, because
(as you said) it makes no difference as long as everything is cached.
I don't believe that the loopback mount makes *any* difference - the dentry
cache is the same.
Perhaps on a loaded system sharding would help, because a lot of dentries
could be evicted - on a single directory that could be much harder.
> > - We'll create shards of 1000 entries each. Anyone who has a repository
> > that's larger than a million revisions will likely be running on a
> > decent filesystem. Even if they aren't, we will still be much better
> > than we were before.
>
> I concur that this is a reasonable course. Malcolm++
+1 to both of you ;-)
Malcolm, thank you!
Regards,
Phil
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Tue Mar 20 15:38:19 2007