Michelle Konzack writes:

I have setup following machines:

1)  PostgreSQL (for authentication)
2)  courier-mta (in/out bound)
    running in the same time spamassassin and clamav-ng
3)  courier-imap-proxy
4)  courier-imap (mailbox storage)
5)  internal DNS server

I have over 1800 accounts stored on 4) and the machines are nearly  idle
even if I it get more  then  320.000  messages  a  day  plus  more  then
700.000 spams which where rejected to 95% on SMTP level.

My question  is,  if  such  machines  are  nearly  sleeping  how  can  I
determiner how big a machine must be?

You might want to double-check your metrics. spamassassin and clam-av, or any other content scanner for that matter, is very CPU intensive. It would be very unusual not to notice any load from that.

That I can add a bunch of storage machines 4) is clear and I  will  need
it for my own network, but what performance do I need for the proxy?

The proxy will need to handle exactly double the sum total of I/O bandwidth of the storage servers. It is proxying all TCP traffic. Everything that gets sent and received by each storage server goes in and out of the proxy server. Every packet sent by the storage server gets received and resent by the proxy server. Every packet received by the storage server gets received and resent by the proxy server.

Note that you can scale the proxy servers horizontally the same way you're scaling out the storage servers. If a single proxy server can't handle the bandwidth, you can just add another with the same configuration, and load-balance them in DNS.


Attachment: pgpiZ3czHpaAH.pgp
Description: PGP signature

------------------------------------------------------------------------------
_______________________________________________
courier-users mailing list
[email protected]
Unsubscribe: https://lists.sourceforge.net/lists/listinfo/courier-users

Reply via email to