Monday, October 25, 2004
I dunno, maybe I am barking up the wrong tree here. I'm in Orlando this week at the Exchange Connections Conference. Love the weather! I was at a session this morning and something really got my goat. Of course, this isn't an official Microsoft Conference or anything, but I attended a session this morning about Exchange Performance Tuning and the speaker made reference to several things that I disagree with.
1. A recommendation was made to go ahead and use the additional capability of mailbox stores and storage groups. Ok - I can see this making sense, but certainly not for everyone. This is the type of recommendation that can really throw some companies over the edge in terms of their server performance. I felt like this comment was unwarranted and irresponsible. Sure the capability exists to use multiple storage groups and multiple databases, but caution needs to be exercised when doing so, otherwise you may well end up with a server that performs worse than before. What am I talking about? Well, for starters, you need to have dedicated spindles for each set of transaction logs. There's more than one set, you say? Yup. Each new storage group has it's own set of Transaction logs. If you use all 4 available storage groups, and keep all the transaction logs on one volume, that would get pretty messy. With the type of audience that normally attends these types of conferences varying widely in level of experience, I doubt that everyone knew this. Not to mention that each additional storage group and each additional database increases the memory requirements to startup Exchange. On a properly built server, this isn't an issue, and of course every Exchange server is always properly built (smirk)
2. Another recommendation was made to fight for the right to be able to do Exchange maintenance. More specifically, being able to run offline defrag/compaction of the databases on a periodic basis. This I also did not agree with and had to make a concerted effort not to make a comment (I felt it would have been out of place in a large group like that). I have been asking around and I have not been able to find any documentation (as of yet) that truly shows me the performance gains from performing routine offline compaction of Exchange databases. Sure, it clears out the white space, but unless you have implemented mailbox limits or moved a ton of mailboxes off of the store, you are actually making Exchange work harder (IMHO)because it has to create new space for the growing database again instead of re-using the existing white space. This to me seems more costly than dealing with a bit of a fragmentation issue. I'd welcome to chance to see some actual numbers in terms of performance before/after, etc, but I suspect these numbers may vary widely depending on the server.
On a side note, I quite enjoyed the keynote last night from Mark Minasi. We all had quite a few chuckles. Dave Thomspon (Exchange VP at Microsoft) also spoke and shared some good material.
I'll blog again later after attending more sessions.