[Linux-cluster] GFS and server performance = Application

[isplist at logicore.net]

Thank you! All good leads. I'll look into all of these.

On Thu, 29 Nov 2007 09:50:15 -0500, Alexandre Racine wrote:
> Performance links
> 
> RedHat Performance FAQ http://kbase.redhat.com/faq/FAQ_78_3152.shtm
> 
> GFS Performance Tuning http://sourceware.org/cluster/faq.html#gfs_tuning
> 
> Mount with noatime http://man.chinaunix.net/linux/redhat/rh-gfs-en-6.0/s1-
> manage-atimeconf.html
> 
> Turn off disk quotas http://www.redhat.com/archives/linux-cluster/2006-
> August/msg00237.html
> 
> Alexandre Racine
> Projets spéciaux
> 514-461-1300 poste 3304
> alexandre.racine at mhicc.org
> 
> 
> -----Original Message-----
> From: linux-cluster-bounces at redhat.com on behalf of isplist at logicore.net
> Sent: Wed 2007-11-28 21:08
> To: linux-cluster
> Subject: [Linux-cluster] GFS and server performance = Application
> 
> I've been trying to get a handle on web server performance on GFS mounted
> storage vs none. Since the other thread kind of got lost, I decided to
> start a
> new one. What I've found is interesting enough that I felt I should talk
> about
> it since I'm surely not the only one using Joomla, in this case.
> 
> This test is without GFS mounted as the root of the web server. The root of
> the site only has an index.html file with little in it;
> 
> #ab -k -n 100 -c 100 http://192.168.1.92/
> 
> Time taken for tests:   0.279518 seconds
> Requests per second:    357.76 [#/sec] (mean)
> Time per request:       279.518 [ms] (mean)
> Time per request:       2.795 [ms] (mean, across all concurrent requests)
> Transfer rate:          146.68 [Kbytes/sec] received
> 
> Same test with GFS mounted as the root of the web server. The root of the
> site
> only has an index.html file with little in it;
> 
> # ab -k -n 100 -c 100 http://192.168.1.92/
> 
> Time taken for tests:   0.162151 seconds
> Requests per second:    616.71 [#/sec] (mean)
> Time per request:       162.151 [ms] (mean)
> Time per request:       1.622 [ms] (mean, across all concurrent requests)
> Transfer rate:          160.34 [Kbytes/sec] received
> 
> Similar right? Now, let's try the same test but this time, we add a full
> bore
> application, a Joomla site in this case at the root of the web server;
> 
> #ab -k -n 100 -c 100 http://192.168.1.92/
> 
> Time taken for tests:   33.583784 seconds
> Requests per second:    2.98 [#/sec] (mean)
> Time per request:       33583.782 [ms] (mean)
> Time per request:       335.838 [ms] (mean, across all concurrent requests)
> Transfer rate:          27.42 [Kbytes/sec] received
> 
> Quite the difference and little to do with GFS from what I can tell. And,
> this
> is what I am trying to confirm, and am asking from the community. Is there
> any
> fine tuning needed for GFS and the cluster itself as well as what ever I
> will
> do with the cluster later?
> 
> In another thread, I was told that GFS would hurt performance and my thought
> was that well, yes, it would take up some of the servers resources but there
> should be plenty left over to handle web serving or what ever else the
> server
> needs to serve up.
> 
> I've tested this in various ways today, from external connections, internal,
> various httpd.conf settings, it's always the same. While some of the
> httpd.conf settings have some effects, the biggest one is always what
> application is being run on the server. In this case, Joomla seems to be
> insanely resource intensive.
> 
> Are there any thoughts on this so that I can know where I need to spend my
> time now. Should I worry about GFS and the cluster itself or move on and
> start
> trying to figure out how to get Joomla to run more effectively?
> 
> Mike
> 
> 
> --
> Linux-cluster mailing list
> Linux-cluster at redhat.com
> https://www.redhat.com/mailman/listinfo/linux-cluster