For many IT shops, management of block-based storage virtualization projects from the initial implementation through ongoing maintenance will require the same level of effort they put into deploying a SAN for the first time, with project complexity dependent on a variety of factors.
IT organizations often adopt storage virtualization technology in conjunction with the replacement of old hardware, the purchase of new storage systems or a shift from one brand of storage to another. Some are deploying networked storage for the first time.
The planning steps are often "no different than they are with any storage project," said Marc Staimer, president of Dragon Slayer Consulting. "You've just added a layer in between the application and the storage system."
Storage virtualization can prove helpful with migrating data because the technology creates an abstraction layer, allowing users to view and manage multiple storage systems as a single pool. Storage virtualization products such as IBM's SAN Volume Controller (SVC) and Hitachi Data Systems Corp.'s USP V claim to virtualize different types and brands of storage systems, making them potentially useful to companies that have undergone a merger or acquisition. IDC also tracks a class of "embedded virtualization" products that can virtualize only their own storage. Vendors in this category include 3PAR Inc., Compellent Technologies Inc., Dell Inc. (with EqualLogic) and Hewlett-Packard (HP) Co. (with its StorageWorks P4000, which it acquired from LeftHand Networks Inc.).
Implementing storage virtualization: Elements to consider
When implementing storage virtualization, IT shops have additional elements to consider depending on what they're hoping to achieve. This is the case whether the technology is in-band or out-of-band, or if the engine is in the server, the switch, an appliance or an array.
"If you are doing in-line or in-band storage virtualization, you would probably have to re-cable. Most storage systems have two redundant connections, so there would be a switch-over phase where you put this in and move things across," said Valdis Filks, a research director at Gartner Inc. Some software-based storage virtualization may require users to unmount the file system, install the software and then remount, he added.
"When people hear the words 'storage virtualization,' their expectations are they can do everything and anything from the virtualization layer, but they cannot," he cautioned. "Storage virtualization does not replace device configuration and device management."
Filks recommends that users consult with the professional services team for whichever vendor's product they choose before undertaking any storage virtualization project. "There is definitely a transition phase and a lot of planning. It's not something you can do easily," he said.
IBM's SVC, one of the most popular storage virtualization products, is an appliance that sits in the data path and is generally installed in pairs, in an active-active configuration, so there's no single point of failure, in keeping with general best practices. But those who want to use the product with existing storage will soon discover there's a considerable amount of upfront work to configure their back-end storage devices.
"When you initially install SAN Volume Controller," said Glenda Fuller, a managing consultant with storage lab services at IBM, "one of the things that most customers need to realize is you actually have to get the current existing storage managed by SVC before you actually get to take advantage of all these wonderful things that SVC does. So, there is a lot of migration that needs to occur."
Fuller said she typically prefers to have a "clean environment" in order to lay out the back end for optimal performance. "It's easier," she said, "if the back-end storage is not encumbered with data on it already."
Although users can virtualize and manage heterogeneous storage systems with SVC, they tend to migrate to IBM storage over time, noted Chris Saul, marketing manager for SVC at IBM. Only after a user migrates data into SVC can they reap the benefits of migrating data nondisruptively to faster, cheaper or different storage, he said.
"You pay the price of configuring the storage once with virtualization, and then after that, the task gets a lot simpler," said Steve Correl, a managing consultant and storage architect at IBM.
Once the users put in all that effort, they may not want to ever back out.
"It is quite a big commitment that you make when you decide to virtualize storage, in much the same way as you make a big commitment when you choose to virtualize servers," Saul acknowledged.
Although IBM claims to have customers using its SVC appliance with heterogeneous storage, Dragon Slayer Consulting's Staimer said many users don't go that route because of the potential for finger-pointing in the event a problem arises.
"Getting support is really difficult," he said.
Staimer added that appliance-based storage virtualization can also be more of hassle to implement and maintain than the array-based options.
"You're adding a layer of management in there. That should be transparent to an administrator, but when troubleshooting occurs, it's not so transparent," Staimer said. "Troubleshooting is the biggest bugaboo with the appliance because now you've got to figure out if [the problem] is before the appliance or after the appliance. Once you put a virtualization layer in, it's hard to see behind the curtain."
This was first published in June 2010