Do You Really Need a SAN Anymore?

12:30 PM -- Forrester Research Inc. analyst Andrew Reichman asks the headlined question in a new report that's sure to ruffle the feathers of a few storage administrators, not to mention vendors. While I feel the same about my SAN as Charlton Heston did about his guns -- they can pry it from my cold, dead hands -- Reichman does have an interesting point to make about moving storage intelligence from the array to the application.

After lamenting how SAN technology hasn't led us to the Holy Land of 80 percent utilization and reduced costs that vendors promised, Reichman posits a new storage architecture for the enterprise with application-based silos of clustered DAS.

I, and most vendors, would argue the reasons for poor utilization are much more human than technological. I've certainly gotten requests from DBAs for 4 TBs of space for an application that could need that much four years from now -- if it's widely adopted across multiple divisions that haven't even seen it yet. I know that wouldn't be any different if we didn't have a SAN.

His real point, and one I could agree with and even argue further, is that application developers have a better handle on their data than a block storage device ever could.

VMware snapshots solve many, if not all, of the problems that array-based snapshots do, and Exchange 2007's CCR and SCR replication features can easily provide DR capabilities for everyone's favorite email server. After all, if you use a block- or file-based replication system for a journaling database like Exchange, Oracle, or SQL server, you have to replicate both the database and transaction logs. Replicate transactions themselves, and you've not only reduced the amount of data 50 percent, you've also eliminated replication traffic events like database defragmentation.Some food for thought -- but not a reason to throw out my SAN. Our Exchange 2007 servers use an iSCSI SAN and disk array, but they use SCR to replicate to our DR site.

Those who want to know about Forrester's arguments can download the 16 page report for a fee.

Howard Marks is chief scientist at Networks Are Our Lives Inc., a Hoboken, N.J.-based consultancy where he's been beating storage network systems into submission and writing about it in computer magazines since 1987. He currently writes for InformationWeek, which is published by the same company as Byte and Switch.