On Tue, Jul 3, 2012 at 11:37 AM, Charles Marcus <cmar...@media-brokers.com> wrote: > On 2012-07-03 3:12 AM, Kaya Saman <kayasa...@gmail.com> wrote: >> >> However this is a clean server with plenty of space left on the pool >> allocated for mail and it's additionally using ZFS too. > > > What OS? ZFS implementation/version? How is mail stored (maildir? mbox?) > > While I don't think this is your problem, just fyi, my understanding is that > it is fairly easy to implement ZFS wrong (which would cause serious > performance problems), and that the only decent ZFS implementation is Suns > (ie, what ships with Nexenta), or the latest FreeBSDs... > > Also, my understanding is that ZFS isn't the snappiest of filesystems even > when properly configured (you trade performance for data integrity). > > Personally, I'd recommend trying this on a traditional FS (XFS or Reiserfs > for maildir) and see if that changes things.
FreeBSD 8.2 x64 using Maildir. ZFS is perfect no worries with that!!! Additionally the system is on a VMware cluster which is also fine - have checked all as diagnostics. The usage here is minimal, and since I also use ZFS at home too with quite a larger file system then at work (I know I know) and really hammer the heck out of it there is no issue. > > > On 2012-07-03 3:12 AM, Kaya Saman <kayasa...@gmail.com> wrote: >> The point is that I am monitoring using nload as well as other things >> and the maximum bandwidth being got with Outlook is a few Mbps burst, >> average 50kbps; while with T-Bird I get way over 130Mbps? > > Congrats - there's your problem... now you need to find out *why* this is so > slow... most likely a tcp dump analysis of a session is the only way - I > think there are people here who could help you analyze one (but not me, > sorry)... > Yeah, it seems to be M$ implementation of IMAP. I don't think that there's anything anyone can do.... Outlook seems to wait after each transmission (found using Wireshark). > > On 2012-07-03 3:41 AM, Kaya Saman <kayasa...@gmail.com> wrote: >> The PST's seem to be stored on local hard disk too. > > 'Seem' to be? You need to make sure, because if they aren't that could > definitely cause, or at least contribute to this kind of problem. > It is definitely stored locally! > -- > > Best regards, > > Charles Regards, Kaya
