What is the Real Effect of Removing Single Instance Storage in Exchange Server 2010

When Exchange Server 2010 was released a lot was made of the decision by Microsoft to remove Single Instance Storage (SIS) from the Exchange database engine.

Plenty of articles have been written about this so I won’t revisit the issue in much detail, except to summarise with these points:

• Disk storage technology has basically plateaued in speed capabilities (ie, disks aren’t getting faster)
• Disk storage technology is a lot cheaper in high capacity, lower performance types (eg SATA II, SAS)
• The previous Exchange database engine used a schema that permitted SIS but did not permit further optimization of performance (specifically, allowing sequential reads)
• Removing SIS and redesigning the database schema to permit sequential reads resulted in massive performance improvements (as much as 70% less disk IO for typical behaviour)

Does This Mean More Exchange Storage?

The major concern from Exchange Server customers was the impact this would have on their Exchange storage requirements.  If SIS is removed, does that mean more disk and tape storage is going to be required?  Especially as more and more organizations are already moving to larger mailboxes?

Furthermore, is that increase going to be exponential because of the increasing popularity of disk duplication (eg SAN mirroring) and the nature of Exchange Server 2010 Database Availability Groups (multiple database copies across many servers)?

Short answer, yes.  All of those storage requirements are likely to increase when your organization moves to Exchange Server 2010.

How Much More Storage is Needed?

In real world migrations I have seen mailbox databases grow by between 20% and 50% just from moving all of the mailboxes from an Exchange 2003 or 2007 server to Exchange Server 2010.

Similarly, growth of email storage over time also increases by similar factors.

So with that in mind, how can an organization mitigate the risk of storage costs getting out of control when they move to Exchange Server 2010?

Reducing Storage Costs for Exchange Server 2010

Firstly, take advantage of the Exchange Server 2010 database performance improvements by deploying Exchange 2010 on lower cost storage (eg SATA II or SAS instead of 15k SCSI).  Some customers are tempted to use “what we’ve already got” and deploy on their existing high performance SAN, when in reality a smarter move would be to provision the Exchange 2010 mailbox servers with lower cost, lower speed direct-attached storage (DAS) for storing mailbox databases.

Secondly, don’t duplicate Exchange 2010 data unnecessarily.  If you have deployed an Exchange 2010 DAG, don’t also utilize SAN mirroring for mailbox database storage.  Let the Exchange 2010 application-layer replication handle it for you (Exchange 2010 SP1 introduced block-level replication to resolve one of the remaining criticisms of the Exchange 2010 asynchronous file-level replication).

Finally, look to alternative methods of de-duplicating Exchange mailbox data in organizations utilizing large mailboxes.  For example, many backup applications are now including data de-dupe capabilities, as do enterprise-grade email archiving solutions.