[RCD] another performance fix
ziba at umich.edu
Fri Dec 5 18:13:01 CET 2008
Dennis P. Nikolaenko wrote:
> Ziba Scott wrote:
>> till wrote:
>>> On Sat, Nov 29, 2008 at 1:03 PM, Thomas Bruederli
>>> <roundcube at gmail.com> wrote:
>>>> till wrote:
>>>>> I agree that we should keep this redirect.
>>>>> @alec If you want to avoid connections at all cost, did you try
>>>>> running an IMAP proxy to see if you can gain some there?
>>> Just found the link:
>> We use imapproxy in production for both HORDE/IMP and RoundCube. We
>> have a proxy processes on each of our webmail servers which have
>> permission to connect to IMAP as any user. Once the webmail client
>> connects to the proxy, the proxy will hold a connection open to the imap
> Have you made any performance measurements? How many users do you
> have? How many users are simultaneously online? How many connections
> are between imapproxy and IMAP server at the same time on average and
> at maximum?
We haven't done any measurements on roundcube without imapproxy, and I
don't have the numbers for concurrent proxy<->imap connections (I'm
looking into that).
I do have a few metrics on hand. Here's some stats for an 24 hour
period in October:
Total webmail hits: 13,446,857
Total HORDE hits: 8,958,309
Total RoundCube hits: 4,214,495
Unique Users: 32,326
HORDE Users: 27,930
RoundCube Users: 12,397
We're running about 15 webmail machines in round robin dns (which
requires a patch to hold outgoing attachments in mysql).
Our peak day for RoundCube had 23,858 unique users and 5,736,260 hits.
Speed has generally been good to acceptable with some instances where
cache issues pulled performance down to a painful crawl.
(I'm neck deep in a cache rewrite if anyone cares).
University of Michigan
List info: http://lists.roundcube.net/dev/
More information about the Dev