Palm Desert, Calif. -- Storage networking technology is being pitched as the rescue vehicle for companies in danger of being buried under an avalanche of data. But experienced users warned at a Computerworld-sponsored conference here last week that building a networked storage architecture can cause rumbling in its own right.

Before a storage-area network architecture can be built to handle all of a company's data, existing IT resources have to be accounted for in order to provide an accurate starting view, noted Stevan Arbona, a consulting project leader at The Goldman Sachs Group in New York. "If you can't measure a process, you can't manage it," he said.

To help cope with the doubling of its data annually, Goldman Sachs is considering a move away from locally attached storage into a networked setup supporting dynamic volume distribution, management and sizing. Included in the plan is the consolidation of more than 700 servers into a smaller number of systems.

Until it conducted an inventory of its storage capacity, the investment banking firm had no hard figures on how much storage capacity it had and how much of that was actually being used, Arbona said. The inventory found that only 52 percent of the capacity was in use, resulting in a large surplus that was being stockpiled because IT managers weren't sure how much space was needed.

And things can become even more complicated once a SAN implementation gets under way.

Michael Butler, a vice president at Morgan Stanley Dean Witter & Co., said the New York-based brokerage faced initial challenges that included incompatible switches from multiple vendors, a lack of robust network management software and inadequate technical support.

"To solve these problems, we basically did it ourselves," Butler said.

Morgan Stanley, which began work on its SAN project 18 months ago, created an internal integration team as well as storage engineering groups at each office campus to perform trend analysis and capacity planning and install new storage technology without affecting the company's existing systems.

Butler said the company wanted to create a more reliable architecture for its 250TB of storage capacity on 4,000 Windows NT and Unix servers. At the same time, it wanted to increase its storage utilization rates, which ranged from 35 to 50 percent. Data availability, disaster recovery and usage have been improved, he added, but there still are issues to be resolved.

Morgan Stanley's approach so far has been to install numerous separate SANs. For example, Butler said, the company has five SANs in the New York area. Each SAN fabric is duplicated and linked to its backup through fiber-optic rings with dense wave division multiplexing. The current setup is a "tactical solution" that's reliable but not very scalable, he said.

Interoperability is also a key issue, said managers. Like other users at the conference, Eric Dryden, an information systems project manager at Houston-based Agip Petroleum Co., said he was pleasantly surprised to hear rival storage vendors at least talking about working together to create interoperability standards that could make SAN implementations easier.

"In the outside world, when you talk to [a particular vendor], all you get is their product," Dryden said. As he browsed through the conference's interoperability lab, Dryden lamented his inability to find technology that could be used to consolidate Agip's tape backup systems.

One user said the best way to deal with interoperability is to avoid it altogether.

Gary Fox, a senior vice president and director of enterprise data storage at Charlotte, N.C.-based financial services company First Union Corp., advised attendees to steer clear of mixed vendor environments for the time being. Users should keep their storage networking architectures as simple as possible by using homogeneous switches, he said.

Computerworld

• Email
• Print
• Share
  • Slashdot
  • Digg
  • Stumble
  • Reddit
  • Newsvine