Four years ago, Jim Wicker had a computing mess on his hands. An acquisition binge by his company, Dynamex—a Dallas-based provider of same-day delivery services—had left Wicker managing an information network with 38 different computing systems, 12 operating platforms and six accounting applications.
Wicker and his team embarked on a step-by-step program to consolidate servers and simplify data-center operations. Within two years, they had decided to deploy a $600,000 storage area network.
Still a relatively novel concept at the time, a storage area network (SAN) is a high-speed network that links storage systems and servers, using switches to route requested data where it is needed.
“The SAN gives us the ability to manage our storage pool intelligently,” says Wicker, vice president of information services for the company, which processes more than 40,000 on-demand transactions a day via its computer network. “The real value it has created for us is in reduced cost of ownership.” He says the company now manages its computing infrastructure with 18 workers—down from 52 when the infrastructure was at its most complex. Only one of those employees oversees the company’s half-terabyte of data.
The very fact that a SAN is in use at Dynamex—a company with $240 million in annual sales—raises the possibility that acceptance of the SAN may be widening. To date at least, SANs have been used primarily by Fortune 500 companies, which have the resources to deal with the evolving technology and with inter-vendor compatibility problems that remain an issue. But as Fortune 500 technology purchasing has fallen off, SAN vendors have been forced to broaden their target market. For vendors like EMC, Storage Technology, and Hitachi (profiled on pp. 77, 78 and 80 respectively), midsized companies like Dynamex represent an emerging opportunity.
SAN equipment has grown rapidly in the last few years and now constitutes a $5 billion-$7 billion market. In addition to Hitachi, EMC and Storage Technology, the hardware equipment makers also include IBM, Hewlett-Packard, Sun and Dell. Then there are specialized network equipment companies like Brocade and Nishan and software makers like Veritas, Qlogic and BMC.
These companies would not have a market if SANs hadn’t proven to be a demonstrably good investment. But they have, especially for the big organizations that have historically deployed them.
Take BlueCross BlueShield of Tennessee. A year after spending $100,000 on SAN equipment to manage a 50-terabyte data warehouse, the insurer claims its investment has already saved it $1.5 million. Part of the savings has come from trimming the number of workers devoted full time to storage management: just one worker now, compared to two previously.
The bigger savings comes from the ability the Chattanooga insurer now has, to manage network resources more intelligently. Information gathered by the SAN is used to identify which pieces of the network are being used most, making it easier to spend money only on systems that need upgrading, says Bob Venable, director of enterprise storage.
In one case, he says, an administrator of a Windows NT-based database at the company had assumed his application was running near its maximum capacity. Management and tracking tools that came with the SAN, however, made it clear the application was using only 25% of its potential resources.
“Without the SAN, we may have gone ahead and purchased another disc-server platform for that application,” Venable says. “Now, that administrator has a feeling of confidence that he still has the room to grow by a factor of four.”
Indeed, the metrics of SAN deployment are often so clear that many companies don’t make the investment unless they can pinpoint not only what the payback will be but when it will happen. Carlson Companies Inc. is one such company; the $30 billion owner of such properties as Radisson Hotels and T.G.I. Friday’s makes deployment of its SANs contingent on savings that can be identified up front.
Carlson is investing more than $1.5 million in SAN equipment and associated switching infrastructure supplied by Hewlett-Packard and Nishan. Gary Johnson, architectural consultant with Carlson Shared Services, says the SAN will allow Carlson to consolidate five servers housing 10 terabytes of data into a single larger server. Johnson says the savings on maintenance costs alone will produce a payback in less than three years.
For all their benefits, though, the adoption of SANs is now being slowed by technological uncertainty. SANs are based on Fibre Channel, a networking technology that is extremely fast but that is also expensive, and which vendors implement in noncompatible ways. Those problems have opened the door to an alternative called iSCSI, which can transport data at Fibre Channel-like speeds, but do so across a standardized TCP/IP network. With the deeper pool of experienced TCP/IP talent that exists, some enterprise buyers may be content to wait until 2003 or 2004 and deploy an iSCSI system as questions over iSCSI standards, interoperability and reliability get addressed. (For a primer on storage architectures, see Baseline, January/February 2002, or find the article online at www.baselinemag.com.)
Back at Dynamex, Wicker has no doubt of the value the SAN delivers for his company. Not only has the system reduced the cost of ownership associated with maintaining the company’s data systems, it also has laid the groundwork for future data management initiatives. Next on the list is the development of a mirrored storage capability in which the SAN will be used to produce a complete copy of all of the corporation’s data that is stored separately from the databases used in the company’s day-to-day operations.
“The reason we can cost-justify this is that it provides tangible cost savings,” Wicker says. “But it’s nice to have that security blanket there, too.”
IBM technical paper on SANs
www.baselinemag.com/sanpaper
IBM’s storage-area-network “redbook”—one of many technical manuals produced by the company’s support organization—offers a detailed overview of SANs and the advantages of implementing them. Only downside is its inherent bias toward IBM.
