I frequently consult with customers who need to figure out how to better store, consolidate, protect, and manage their SQL databases, Exchange servers, and MOSS farms.
With Microsoft, most of the time I am simply in awe for their long-term ability to seed and harvest a market by slowing slipping innovations into gradually more expensive and higher-quality V2 and V3 products. SQL Server, Zune, SCOM, and Hyper-V are all illustrations of this. But sometimes you get a rogue product group within the company that acts without regards for their overall company strategy. And maybe that’s what happens when you have 65% market share.
I am speaking specifically about how people are being swayed away from using “expensive SAN hardware” for storing and protecting Exchange 2007 data. Not only do the advocates of this approach (which includes the Exchange product team) obscure long term costs by comparing pure “cheap storage” or “DAS” acquisition costs and ignore long-term cost of ownership, they also ignore:
- Management and simplicity of shared storage devices (one place to put stuff)
- Utilization benefits of shared storage device (one place storage means more likely to use it)
- Bandwidth requirements of the CCR/SCR replication scheme which can be 5 times larger than cached replication appliances (say EMC’s RecoverPoint)
- Exchange server virtualization – massive amounts of people are virtualizing their Exchange servers and placing data on a SAN to get the maximum benefit (virtual servers need virtualized/SAN storage to enable most of the advanced features).
Paul Galjan writes a very convincing blog post about dissonance within Microsoft, their latest anti-SAN calculator, and how this anti-SAN stance really backfires when most of the company is aggressively pursuing a very heavy virtualization strategy and taking a more balanced approach about which virtualization platform to use (like ESX being certified under SVVP). I travel around the country giving workshops on how to virtualize Exchange and it’s truly up to the customer to decide whether I talk about Hyper-V or VMware ESX as their hypervisor of choice. At the end of the day, the customer will guide that decision. Not me. I am not paid to push VMware.
If it’s really a cost issue, perhaps customers could chain a bunch of USB drives together and create a super-DAS configuration. But wait. That is pretty close to what they are saying, isn’t it? Get rid of the SAN, get a lot more servers and some cheap disk, and prepare to throw a lot of people at the new storage management problems that will be sure to take place because you aren’t sharing storage across servers, and you’re not sharing global hot spares with hundreds of disks in a system, or taking advantage of a nice big cache layer that smooths out most of the peaks during the Outlook users workday. And not taking advantage of high speed LAN-free backup snapshots or clones for recovery from corruption.
I’ve seen customers trade in SANs and array-based replication schemes for this new model, and only months later come back to an EMC SAN (and replication) saying “OK, you were right… Just don’t say I told you so.”
But it’s not like me to say that to my customers. I warn them up front. :)
Some interesting data I’ve been able to gather recently: