[horde] Hardware/config recommendations for large rollout

[Ron Hall]

>My University handles a large service base by putting lots of servers
>behind a load balancer.  It's not so important what kind of machines
>you use, as long as you use enough. The more powerful the machines,
>the fewer you need, and vise-versa.
>
>  
>
    I would like to echo this sentiment - we currently have 4 machines 
behind a load balancer
    All running the same OS - 3 are Xeon and the 4th something else. 
There is s dedicated DB
    backend and if I were redoing it today I would consider some 
clustering technology.

    The nice thing about the multiple heads is that we can seamlessly 
remove one from the mix
    for maintenance. Currently we can cycle full OS/Software install on 
all the boxes in two days
    without any noticable interuption in service. (these a 7 hour days - 
we could do it in a single
    day if we worked after hours :) ).

    The other part of our puzzle was the php-mmcache - we find that 
helps alot. We have never been
    able to get an imap-proxy working properly, but I suspect we could 
gain some advantages there.

    Our original operational goal was to service > 2500 simultaneous 
connections - took us about 2 weeks
    of hell to settle on the right parameters and configuration, but we 
got it there - works reasonably well
    few complaints.

    r