[horde] Apache Performance tuning with Horde

[Michael M Slusarz]

Quoting Arjen de Korte <arjen+horde at de-korte.org>:

> On a single machine, filebased caching using a tmpfs volume will be  
> faster than Memcached. Unless you're using multiple servers in  
> parallel, I'd recommend against using Memcached.

For a Horde server running on a single server, this is true for ANY  
caching/session backend, not just memcached.

As we (hopefully) have made clear in the documentation: there is zero  
reason to use anything other than file-based storage backends (except  
for backends requiring relational queries, e.g. prefs) for a  
single-host server.

>> Is there a possibility to speed up things? Any caching on IMAP-side?
>
> If you don't already have that, this will probably yield the most.
>
>> Is this behavior just because we are using a single processor environment?
>
> Only if you're not limited by I/O, which you really should be  
> checking by profiling what is happening. Xdebug and cachegrind (I  
> use KCacheGrind) are your friends here. You wouldn't be the first to  
> find that your PHP processes are spending a lot of time in usleep().  
> See http://www.xdebug.org/docs/profiler for more info.

In current versions of Horde, there is essentially nothing that should  
be causing a CPU spike for more than a second.  If so, you have likely  
found a bug.  In this case, there's a good chance that there is not a  
bug in our code but rather the way an external device has implemented  
a protocol (ActiveSync).

michael

___________________________________
Michael Slusarz [slusarz at horde.org]