Practical Analysis: Avoid The Direct-Attached Storage Trap

InformationWeek SMB digital issue - June, 2011

InformationWeek Green

Art Wittmann

Each year, InformationWeek Analytics conducts dozens of surveys, most of which contain a number of questions we've asked for years so that we can find trends in the responses. One question that always interests me is how much storage is direct-attached versus NAS versus SAN. Even in our recent 2011 State of Enterprise Storage Survey, 39% of respondents said that 25% or more of their storage resources are direct-attached. For all but the smallest IT shops, direct-attached storage is a problem just waiting to happen. In fact, unless you're pretty new to IT, it's probably already been a problem for you.

It doesn't take too many 3 a.m. nights sitting in a server room with a pile of CDs and tapes, trying to rebuild a server before work starts the next morning, to have you longing for a better way. And even if you haven't had a disaster, data almost always lives longer than a typical server does, so decoupling data storage from servers makes good sense.

Until recently, server makers weren't much help in encouraging centralized storage. Servers with more memory capacity and CPUs were physically bigger, too--with lots of empty drive bays tempting you to load them up with comparatively cheap storage. There was no altruistic intent on the part of manufacturers. Profit margins on 1U and 2U servers are razor thin, so one way for them to gin up some extra dough has been to add sheet metal, and power and cooling capacity.

Because it's easy to fill those empty drive bays, the cheapest way for buyers to add storage to their networks is to add the direct- attached variety. It's also the hardest storage to manage, the least resistant to failure, and the toughest to fix when a failure does occur. To many IT pros, centralized storage seems even more dangerous: There sits one box with one set of disks that represents a whole lot of eggs in one basket. But good networked storage systems are always built with an N+1 philosophy, with more controllers, fans, power supplies, and drives than would be minimally needed. Even power connects can be redundant so that a SAN or NAS box isn't sitting on just one breaker.

While it probably doesn't take much convincing to get you to think about centralizing storage for data, it's a bigger leap to start putting boot images and virtual machine files on comparatively expensive centralized storage--leaving local disk completely out of the picture or keeping it around just for swap files and temporary storage. Whichever way you decide to go, it makes a lot of sense (of course) to keep gold masters of system and VM images somewhere besides on the server itself.

When it comes to storage, too many midsize enterprises have gone too far with their "we can do more with less" philosophy. If you're still working to minimize your initial purchase spend for storage, re-evaluate that point of view. You'll save time and resources, provide a better service, and have a more responsive IT team with a better, planned, centralized storage approach.

Art Wittmann is director of InformationWeek Analytics, a portfolio of decision-support tools and analyst reports. You can write to him at [email protected].

To find out more about Art Wittmann, please visit his page.

More than 100 major reports will be released this year. Sign up or upgrade your InformationWeek Analytics membership.