No subject

objects in memory and increasing memory usage. This may be causing portions
of memory to swap out where they may not have before. When retrieving the
asset, it would need to swap it back in and it may have to do it in several
steps in order to locate the asset. This may cause further delay as it may
have to swap other memory out to make space. It would seem likely that a
file based cache may perform better as the filename might provide a more
direct route to the actual cache data. I suspect that the operating system
may also allow searching directories for a filename fairly quickly - modern
file systems are quite efficient these days.
This is really just a guess... I suspect some benchmarking and profiling of
several systems of different configurations may be necessary to determine
what the bottleneck really is.

On Tue, May 26, 2009 at 3:52 PM, <diva at metaverseink.com> wrote:

> I noticed we have GlynnTucker.Cache.dll in the distribution, and this
> wasn't being used. I have no idea what the story is, but since it's
> there, I did one alternative cache module implementation using it, just
> to see what happens. Since everything is modularized now, replacing the
> asset cache implementation is as simple as editing your XXXCommon.ini
> and changing the name:
>
> ;AssetCaching = "CoreAssetCache"
> AssetCaching = "GlynnTuckerAssetCache"
>
> I would still urge that someone figures out what's wrong with
> OpenSim/Framework/Cache.cs, which is the one used in CoreAssetCache.
>
> The GlynnTuckerAssetCache has been tried in the UCI Grid and in some
> sims in OSGrid and at least it's not being an anti-cache. Let's wait a
> few more days to see if it's being a real cache; if so we should see
> declines on the traffic to the asset servers :-)
>
> Arthur Valadares wrote:
> > If we ran a profiling with this cache maybe we could see where the
> > bottle neck is, couldn't we?
> >
> > Mono has simple statistical profiler that shouldn't be too intruding:
> >
> > $ mono --profile=default:stat program.exe
> >
> > If anyone who saw this performance loss could test this, would be
> > interesting to see where it's spending all this extra time.
> >
> > On Tue, 2009-05-26 at 10:57 -0700, diva at metaverseink.com wrote:
> >> Just a quick note to inform everyone that word from osgrid folks
> >> indicates that the asset cache is still borked and still a mystery. When
> >> it is on, CPU usage spikes to unsustainable values (150% with 4 avies in
> >> WP).
> >>
> >> If anyone cares to take a look at it / replace it with something else,
> >> please do.
> >>
> >> OpenSim/Framework/Cache.cs
> >> _______________________________________________
> >> Opensim-dev mailing list
> >> Opensim-dev at lists.berlios.de <mailto:Opensim-dev at lists.berlios.de>
> >> https://lists.berlios.de/mailman/listinfo/opensim-dev
> >
> > ------------------------------------------------------------------------
> >
> > _______________________________________________
> > Opensim-dev mailing list
> > Opensim-dev at lists.berlios.de
> > https://lists.berlios.de/mailman/listinfo/opensim-dev
> _______________________________________________
> Opensim-dev mailing list
> Opensim-dev at lists.berlios.de
> https://lists.berlios.de/mailman/listinfo/opensim-dev
>

--0016364ee01c638db7046ae1d243
Content-Type: text/html; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable