On Wed, Jul 11, 2007 at 05:17:49PM -0300, Eduardo Kienetz wrote: > On 7/11/07, Brian Candler <[EMAIL PROTECTED]> wrote: > >On Wed, Jul 11, 2007 at 04:16:26PM +0200, [EMAIL PROTECTED] > >wrote: > >> >> Hello all, > >> >> which filesystem is the best one for a partition > >> >> containg only maildir++ mailboxes? > >> >> > >> >> We already use ReiserFS, but have a big IO-Wait. > >> >> (SEARCH and ORDER is slow on big IMAP folders) > >> >> > >> >> Is XFS, JFS, ext3 (with journal and h-tree) or ext4 better? > >> >> > >> >> Any recommendations? > >> >> > >> > > >> > I take it you use Linux. Go with ext3. Not the fastest, not the best > >one > >> > theoretically, but THE most stable when you use Linux. > >> OK. But which one would be the fastest (most efficient)? > > > >Don't use ReiserFS. When it breaks, the entire filesystem is toast and > >non-recoverable. That is, it has no reliable equivalent of e2fsck. > > I really don't get what you mean... > Have you ever used reiserfsck ? > reiserfsck --rebuild-tree ?
Yes. We spent the best part of a week trying to recover the filesystem. Everything was lost. This was about a year ago, so maybe the tools are better now. > hmm, this made me think about courier-imap... and my graduation project... > Sam, do you think some courier-imap operations could be parallelised > (via MPI on a cluster) to improve performance? I doubt that would be worthwhile. You get parallelisation for free with courier-imap because typically you have many users connected concurrently, and at any one time several may be performing their own operations, so the workload is divided between CPUs already. What I was trying to say was: if you do a SEARCH operation across a big IMAP folder, and you find it's slow, then that may mean that the folder is not indexed sufficiently. That is, it's an algorithmic issue. The filesystem will have little or no impact on performance because all it's doing is opening and reading each file in turn, for which I suspect most filesystems give almost identical performance, limited by the drive's seek time. If that's true, then code changes to courier-imap would be required to improve performance for those operations. Regards, Brian. ------------------------------------------------------------------------- This SF.net email is sponsored by DB2 Express Download DB2 Express C - the FREE version of DB2 express and take control of your XML. No limits. Just data. Click to get it now. http://sourceforge.net/powerbar/db2/ _______________________________________________ Courier-imap mailing list Courier-imap@lists.sourceforge.net Unsubscribe: https://lists.sourceforge.net/lists/listinfo/courier-imap
