I’ve been doing a little digging today, following a query from a partner company who’re helping out one of their customers with some performance problems on Exchange. Said customer is running Exchange 2000, and has some frankly amazing statistics…
… 1000 or so mailboxes, some of which run to over 20Gb in size, with an average size of nearly 3Gb. To make matters even worse, some users have very large numbers of items in their mailbox folders – 60,000 or more. Oh, and all the users are running Outlook in Online mode (ie not cached).
Now, seasoned Exchange professionals the world over would either be shrugging saying that these kind of horror stories are second nature to them (or fainting at the thought of this one), but it’s not really obvious to the average IT admin *why* this kind of story is bad news.
When I used to work for the Exchange product group (back when I could say I was still moderately technical), I posted on the Exchange Team blog (How does your Exchange garden grow?) with some scary stories about how people unknowingly abused their Exchange systems (like the CEO of a company who had a nice clean inbox with 37 items, totalling just over 100kb in size… but a Deleted Items folder that was 7.4Gb in size with nearly 150,000 items).
Just like it’s easy to get sucked into looking at disk size/capacity when planning big Exchange deployments (in reality, it’s IO performance that counts more than storage space), it’s easy to blame big mailboxes for bad performance when in fact, it could be too many items that cause the trouble.
So what’s too many?
Nicole Allen posted a while back on the EHLO! blog, recommending 2,500-5,000 maximum items in the “critical path” folders (Calendar, Contacts, Inbox, Sent Items) and ideally keep the Inbox to less than 1,000 items. Some more detail on the reasoning behind this comes from the Optimizing Storage for Exchange 2003 whitepaper…
As the number of items in the core Exchange 2003 folders increase, the physical disk cost to perform some tasks will also increase for users of Outlook in online mode. Indexes and searches are performed on the client when using Outlook in cached mode. Sorting your Inbox by size for the first time requires the creation of a new index, which will require many disk I/Os. Future sorts of the Inbox by size will be very inexpensive. There is a static number of indexes that you can have, so folks that often sort their folders in many different ways could exceed this limit and cause additional disk I/O.
One potentially important point here is that any folder, when it gets really big, is going to take longer to process when it fills up with items. Sorting or any other view-related activity will take longer, and even retrieving items out of the folder will slow down (and hammer the server at the same time).
Oh, and be careful with archiving systems which leave stubs behind too – you might have reduced the mailbox size, but performance could still be negatively affected if the folders have lots of items left.