[OpenAFS] Re: Openafs performance with Apache/PHP

Nate Gordon nlgordon@gmail.com
Tue, 14 Aug 2007 13:08:44 -0500

On 8/14/07, Jeffrey Altman <jaltman@secure-endpoints.com> wrote:
> Nate Gordon wrote:
> > So I loaded up ntop and isolated http, dns, ssh, afs, and mysql.  I
> > ran the previous test (ab -k -c <users> -t 120 <url>) again for
> > various numbers of users and here is what I came up with:
> > 1 user: H:29.5M D:1.8M A:4.6K M:19.9M O:1.2M
> > 2 users: H:50.3M D:3.0M A:2.0K M:33.7M O:241.9K
> > 3 users: H:56.1M D:3.3M A:0K M:37.8M O:234.6K
> > 4 users: H:56.2M D:3.3M A:0K M:37.9M O:127.9K
> > 5 users: H:48.0M D:2.9M A:2.0K M:32.4M O:246.3K
> > 8 users: H:23.4M D:1.4M A:2.3K M:15.8M O:135.6K
> > 10 users: H:16.6M D:1.0M A:3.8K M:11.3M O:385.6K
> > 20 users: H:16.7M D:1.0M A:0K M:11.7M O:235.5K
> > 30 users: H:16.8M D:1.0M A:1.9K M:11.9M O:346.2K
> >
> > H - http, D - dns, A - afs, M - mysql, O - other
> >
> > This shows an interesting trend in that there is a peak level of
> > performance, as well as a worst case performance level as well.  It
> > also shows that afs traffic is essentially zero.  This would indicate
> > that my performance bottleneck is in accessing the afs cache.
> How were you measuring the AFS traffic?

I used ntop and defined ports 7000-7010 to be in the group afs.

That is so little data moving
> back and forth your users couldn't possibly be using the AFS cache
> if that's all the traffic there is.  At the very least there would
> still have to be queries for access control, status fetches, server
> probes, etc. even assuming that all of the data the user requires is
> already in the cache.

I was under the impression that all of that data(permissions, status,
etc.) is cached with the meta data.  But even if there was more
traffic outside of my AFS grouping it would have shown up in the
'other' column, and that traffic is pretty constant as well.  If I
should be looking at something else besides those ports let me know
and I will use that instead.  Also, realize that this is a benchmark
and not actual users.  I'm hitting one page repeatedly, which should
in theory isolate the performance to only that involving the local
cache and apache/php.  It should also be noted that I did not reset
the cache between runs of my tests.  Again, I'm trying to isolate the
performance of the afs cache through the kernel interface.

I've run a couple of the tests again with tcpdump monitoring afs
traffic (again ports 7000-7010) and it comes up with zero traffic on
afs.  I tested it by just wandering around the file system in places I
hadn't been and traffic does show up.  But if I do hit a
directory/file I have been in before there is zero traffic from afs.

If I let tcpdump sit there long enough I do eventually see random
call-back and status pings, but they don't seem directly related to
any test I'm running.

-Nathan Gordon

If the database server goes down and there is no code to hear it, does
it really go down?