Around murder : Architecture considerations

[LaurentG]

Hello,
 
after testing Cyrus Aggregator 2.2.8. I'm about to to deploy it for a 
potential 100000 users population.
 
The architecture I plan is the following :
 
2 LVS with heartbeat in front
3 frontends with 5Gb of RAM (authenticating against LDAP)
3 backends with 500 Gb each for mailboxes storage and 3 Gb RAM with 
Reiserfs for mailboxes partition
1Mupdate
 
I've few questions around that :
*Replication/backup* :
    What would be your advice for backing up mailboxes? The solution 
from Cambrige University 
(http://www-uxsup.csx.cam.ac.uk/~dpc22/cyrus/replication.html 
<http://www-uxsup.csx.cam.ac.uk/%7Edpc22/cyrus/replication.html> )seems 
attractive, but hasn't been implemented (yet?) into Cyrus. Today what do 
you use in production?  RAID?  0+1?
 
*Webmail*
I'm about to use Horde2/IMP3, and will store sessions in MySQL settled 
on the mupdate server.
Webmail will probably be installed on each frontend.
Any problem with that?
 
*Sieve*
Any preconized client for sieve? Websieve? Any other? (until 
Horde3.0/Ingo become stable).
 
What else ... any use of proxying IMAP?
 
*Any suggestions*, from your experiences, around that, would be welcome. 
Could probably  be usefull for others planning to deploy Cyrus IMAP, I 
guess.


---
Cyrus Home Page: http://asg.web.cmu.edu/cyrus
Cyrus Wiki/FAQ: http://cyruswiki.andrew.cmu.edu
List Archives/Info: http://asg.web.cmu.edu/cyrus/mailing-list.html