Re: [Dovecot] 1.0(.14) vs 1.1.2 performance

2008-08-06 Thread Thomas Hummel
On Wed, Aug 06, 2008 at 10:30:09AM -0400, Timo Sirainen wrote: > Well, one change is that by default copying messages is now done using > hard links, but you could have done that with v1.0 also by changing a > setting. Yes, that's what I did at the time I was running 1.0.14. -- Thomas Hummel

Re: [Dovecot] 1.0(.14) vs 1.1.2 performance

2008-08-06 Thread Timo Sirainen
On Aug 6, 2008, at 4:56 AM, Thomas Hummel wrote: On Tue, Aug 05, 2008 at 06:40:00PM -0400, Timo Sirainen wrote: Do you use POP3? Yes. I don't have the stats right now but I'd say, on 2500 users, 60% are using IMAP, 40% POP3. But I had the feeling that it was the IMAP processes which were

Re: [Dovecot] 1.0(.14) vs 1.1.2 performance

2008-08-06 Thread Thomas Hummel
On Tue, Aug 05, 2008 at 06:40:00PM -0400, Timo Sirainen wrote: > Do you use POP3? Yes. I don't have the stats right now but I'd say, on 2500 users, 60% are using IMAP, 40% POP3. But I had the feeling that it was the IMAP processes which were causing the load, particulary because some IMAP users

Re: [Dovecot] 1.0(.14) vs 1.1.2 performance

2008-08-05 Thread Timo Sirainen
On Aug 5, 2008, at 9:43 AM, Thomas Hummel wrote: I just switched from dovecot-1.0.14 to dovecot-1.1.2 and noticed a significant drop in the server load average (which has become lower and above all doesn't seem to have unexplained peaks). Indexes are still locally stored and nfs attribute

[Dovecot] 1.0(.14) vs 1.1.2 performance

2008-08-05 Thread Thomas Hummel
Hello Timo, I just switched from dovecot-1.0.14 to dovecot-1.1.2 and noticed a significant drop in the server load average (which has become lower and above all doesn't seem to have unexplained peaks). Indexes are still locally stored and nfs attribute caching is still off). I searched the list a