|
Blades offer many advantages over traditional servers, but they create some unique storage and application issues.
PETER CHAU is the infrastructure architect at North Shore Credit Union (NSCU), a North Vancouver, British Columbia, organization with 44,000 members, 12 branches and a new software banking system that runs on a bunch of Hewlett-Packard (HP) Co. blades.
From Chau's vantage point, his blade infrastructure is finally ready for prime time, approximately two years after NSCU got into the blade market with HP's first iteration of blades, the p-Class. That changed Chau's infrastructure to a smaller footprint with decreased power costs, but it didn't change his life. That came later, with HP's c-Class and its Virtual Connect Enterprise Manager, says Chau. By creating bay-specific I/O profiles with unique MAC addresses, Virtual Connect allows network and storage administrators to establish all LAN and SAN connections during deployment, and lets them avoid having to do it again even if additional servers are deployed or existing ones are changed.
"If the blade on slot 1 fails, all I need to do is replace it with a new one," says Chau. "The back-end connections to the SAN and the network remain intact; it doesn't change. Previous to [Virtual Connect], I would have to engage the SAN admin to re-establish the LUN connections and map to a new WWN."
The technology, he says, has not only made his IT shop more efficient, but freed up some precious time for him to work on new projects. Chau's p-Class blades did leverage a SAN back end but, from an admin perspective, he might as well have been using standard rack-mountable servers.
"With the c-Class and its Virtual Connect Onboard Administrator, we feel it's ready to facilitate our banking system and BI [business intelligence] infrastructure," he notes.
The increasing popularity of blades in the data center has helped fuel serious competition among vendors who are all trying to package, or stuff, the same capabilities of a traditional four-socket rack-mount server into a blade form-factor. However, many users and analysts agree that blades, while just another server form-factor, present unique challenges from a storage management perspective. Managing aggregated I/O for multiple blades, proprietary technology, and the mistaken assumption that blades will automatically lessen your power and cooling costs are among the chief challenges for storage pros considering blade strategies.
James Jancewicz rejected two blade server products. Jancewicz, who took over the storage duties at a large insurance organization 10 years ago, says his firm (which he asked not be named due to company policy) brought in competing HP and IBM Corp. blade centers and dumped a workload on them to see how they would perform. As far as I/O is concerned, they didn't overload the individual blades, he says. "But it was much more difficult to manage a box when you had aggregated I/O for multiple blades running over common interfaces--both fibre and IP," he explains.
|
