[Gluster-users] Performance problems

[Krzysztof Strasburger]

On Tue, Mar 30, 2010 at 02:00:58PM +0200, Bedis 9 wrote:
> Hello,
> 
> I had the same configuration/performance.
> In order to get better performance, I have configured apache as a
> reverse-proxy (mem)cache on the front IP, pointing to the apache
> webserver which listen on the local loop.
> 
> That way, the most accessed files are served from memory. The server
> has 16G of memory ;)
Bedis,
you have a nice, fat server, but this only hides the problem,
instead of solving it. We have basically two problems that need at
least an answer, what causes them?
1. Glusterfs is very slow on some sites, much slower than NFS (networking
or client-side operations?)
2. Glusterfs may eat whole memory of the computer it runs on, even without
enabling any caches, provided that the filesystem it serves contains many
files and you run anything that touches a lot of them (ls -R, du).
I hope that both are related to the same bug/misfeature ;).
Krzysztof