The inconvenient truth: Item counts matter

The conventional wisdom in the Microsoft Exchange administrator base is that mailboxes are getting too big in terms of size, hence the advent of quotas and email archiving solutions to reduce mailbox sizes.

However, many Microsoft Exchange gurus have now realized that the number of items in a mailbox and a folder matters as much in some cases as the size of the mailbox itself.

Bob Spurzem at Ferris Research points this out on his blog.  He refers to a great technical article on Microsoft TechNet on the same topic which includes the following:

Folder contents are stored in a table in the information store database. As the number of items increases, there is a corresponding growth in storage complexity. The storage mechanism for the Exchange store is the Extensible Storage Engine (ESE). ESE uses B+ tree data structures to store records. As the number of records increases, the potential number of disk I/O requests that are required to locate the information and traverse the B+ tree also increases. For more information, see Extensible Storage Engine Architecture.

As the number of items increases, the possibility of any requested data being located on physically adjacent locations on the hard disk is greatly diminished. Therefore, more I/O requests are required.

Martin Tuip, on his Archiving 101 blog, mentioned this issue a few weeks ago as well.

As Martin points out, and as I know from experience, the traditional approach of archiving messages out of Exchange but still replacing them with a small message ("stub") that's a pointer to the original sometimes has challenges.  The fact is that if you replace every 5 MB email with a 1 KB email, you'll cut mailbox size but still retain high item count.  Indeed, the lack of mailbox size will encourage users to store more items in the long run.

There are a number of solutions to this problem - from:

• Pruning stubs over time (how many on-premise products handle it)
• Not using stubs at all and directing users to search instead

There is no perfect answer here, but users should be aware that mailbox size is only one of several issues to consider.