Then; as I said, our experiences have been different. The largest server I used the product on was an Exchange 2000 server running a 14 disk external SCSI array that was split into two 6 spindle sets that were stripped then mirrored, or was it mirrored then stripped, don't recall and I don't think it makes much of a difference.
The system handled 850 to 900 users split over 3 MDB's in one storage group. Mailboxes ranged from less than 100 Mbytes to over 5 Gigs. This may or may not fit your definition of a "small workgroup server with few spindles". After several days of 5 and 6 hour backups, I ran a 3rd party defragmentation tool (NOT that piece of tripe that comes with Windows) and saw my backup times drop about 30 minutes per run and my users stopped complaining about how "Email is slow" for about two weeks. I could give a flying spaghetti monster less if you want to continue the argument. I suppose we can agree to disagree on this, but let's stop calling each other vacuum headed idiots for wanting to run something like a defragmentation utility on the server. And I am not a horse, I don't even play one in SecondLife. John H. Matteson, Jr. Systems Administrator/ITT Systems FOB Orgun-E Afghanistan DSN - 318 431 8001 VoSIP - (308) 431 - 0000 Iridium - 717.633.3823 Roshain - 079 - 736 - 3832 "A man who thinks of himself as belonging to a particular national group in America has not yet become an American. And the man who goes among you to trade upon your nationality is no worthy son to live under the Stars and Stripes." Woodrow Wilson -----Original Message----- From: Neil Johnson [mailto:[EMAIL PROTECTED] Sent: Friday, March 28, 2008 1:26 PM To: MS-Exchange Admin Issues Subject: RE: reclaiming space This has been tested pretty intensively recently and found to behave as expected. The only times I have seen ANY difference at all was on tiny workgroup servers with a couple of disk spindles, and quite large databases. Essentially we only get about 25MB/sec per stream from the backup API (per SG), so as long as your disks can hit that, NTFS defrag wont make any difference to online backups (it may reduce IOPS though). Online maintenance is also a mostly random operation, so that's unlikely to see any improvements. The user experience improvement is likely to be perceived, what was the actual drop in RPC Averaged Latency? The one caveat to the NTFS defrag is if the EDB file was severely fragmented beforehand, i.e., hundreds of thousands of fragments. This can cause high CPU for NTFS read and write operations which can affect IO. However, unless you doing something insane like storing other data on the same LUN's as your EDB it would be unusual to see such fragmentation. You can lead a horse to water, but you cant make it drink... -----Original Message----- From: Matteson, John H Jr USA Mr USA 25th SigBN (ITT) [mailto:[EMAIL PROTECTED] Sent: 28 March 2008 07:05 To: MS-Exchange Admin Issues Subject: RE: reclaiming space Then obviously your experiences have been different from mine. Once I did a physical defrag of the hard disk, I chopped a half hour off the backup time to tape, on line database maintenance completed more quickly and my users perceived faster response times from the server. By the laws of aerodynamics, a bumble bee can't fly, but in reality it flies quite well. John H. Matteson, Jr. Systems Administrator/ITT Systems FOB Orgun-E Afghanistan DSN - 318 431 8001 VoSIP - (308) 431 - 0000 Iridium - 717.633.3823 Roshain - 079 - 736 - 3832 "A man who thinks of himself as belonging to a particular national group in America has not yet become an American. And the man who goes among you to trade upon your nationality is no worthy son to live under the Stars and Stripes." Woodrow Wilson -----Original Message----- From: Neil Johnson [mailto:[EMAIL PROTECTED] Sent: Thursday, March 27, 2008 8:50 PM To: MS-Exchange Admin Issues Subject: RE: reclaiming space [snip]...once you've completed the ESEUTIL defrag, like Perfect Disk 2008. That will help you get the file into a contiguous condition and help your disk performance quite a bit.[snip] Exchange IO is random. Having a contiguous EDB doesn't make any difference, except to streaming backup times and CRC checks, and even then is a miniscule difference. ~ Ninja Email Security with Cloudmark Spam Engine Gets Image Spam ~ ~ http://www.sunbeltsoftware.com/Ninja ~ ~ Ninja Email Security with Cloudmark Spam Engine Gets Image Spam ~ ~ http://www.sunbeltsoftware.com/Ninja ~ ~ Ninja Email Security with Cloudmark Spam Engine Gets Image Spam ~ ~ http://www.sunbeltsoftware.com/Ninja ~ ~ Ninja Email Security with Cloudmark Spam Engine Gets Image Spam ~ ~ http://www.sunbeltsoftware.com/Ninja ~
