Server clustering and virtualization give users a better road to dependability.
When MLT Vacations scrapped its pricey Sun hardware for a cluster of IBM Linux blade servers to support its Oracle database, it expected to see savings. But the company got something even more: a database environment it truly could depend on 24/7.
That was "icing on the cake," says Chris Corona, manager of systems services for the online provider of travel packages, in Edina, Minn.
Emerging clustering and virtualization technologies can bring high-end capabilities to low-priced, standards-based servers that have been pumped up with performance and reliability improvements. In turn, IT managers get a wider choice when it comes to mapping business continuity plans. Many, like Corona, are finding that the tools they've selected to make the data center more dynamic - and, interestingly, to cut costs - are creating a more reliable environment to support critical-business processes. Should a failure occur, applications that aren't tied to the hardware on which they run easily can be moved from one virtual server to another or within a cluster.
"When you thought about business continuity, you used to think replicated servers, replicated data and separate physical machines," says Joe Clabby, an analyst with Summit Strategies. "Now you think resource pool."
At International Truck and Engine, a Warrenville, Ill., manufacturer, IT Manager Barry Naber keeps applications running smoothly using VMware's VMotion. This tool lets users move virtual machines among physical servers.
"We brought VMware in for server consolidation and realized it's much bigger than that. It's also business continuity," Naber says. "With VMotion, if one server within our VMware farm is getting hit more heavily than the others, we can move users over to another server unbeknownst to them and allow the application to function without having to take the server down. Before, end users would have experienced an outage."
International Truck and Engine has about 150 VMware partitions running on less than a dozen servers. The company has avoided about $1 million in hardware costs, Naber says.
The tech choices
While VMware, now an EMC company, largely created the virtualization market for x86 servers, others, such as Microsoft, SWsoft and the open source virtualization project Xen, now offer alternatives. IBM, HP and Sun also provide tools to make virtualization and clustering simpler to deploy in enterprise data centers. And clustering tools aimed at making it easier to tie together disparate systems are available from companies such as Emic Networks, Neverfail Group, Penguin Computing, PolyServe and Qlusters.
"What you'll see is more and more integration of all these types of technologies, which lend themselves to more choice and more capabilities within the IT infrastructure for doing business continuity," says Jim Sangster, director of N1 and availability marketing for Sun. "What typically has only been available on mainframe-class systems - well, now we can do it on a cheap little Opteron box."
The move to Linux and Intel-based systems is expected to save MLT Vacations as much as $1 million over the next five years, Corona says. Maybe more importantly, with the Linux-based cluster, which uses PolyServe on the back end to manage the distribution of data across the server nodes, the company says it hopes eventually to eliminate downtime altogether.
"The way things worked before, if we had an outage in a database server, we were probably looking at anywhere from two to four hours of downtime," Corona says. "Now, as long as there is one node up and running, getting back online takes as long as it does to restart the application, which can be done remotely in about 15 minutes." During the next year, Corona plans to make the reservation application cluster-aware "so it can reconnect on failure," he says.
The need to reconnect upon failure is what led online mail order pharmacy CanadaDrugs.com to clustering on low-cost Linux servers, says Curtis Anderson, IT director at the Winnipeg, Manitoba, company. As a Microsoft user, the company had faced outages weekly. Today, it uses Emic's high-availability clustering software to keep Linux applications up and running.
"We didn't really have a business-continuity plan before we brought in the cluster," Anderson says. Downtime was costing CanadaDrugs.com about $50,000 per week, he says.
"Now that the application is distributed across multiple nodes and clustered, it gives you more options" when it comes to business continuity, Anderson says.
Wayne McDilda, senior security analyst for Mirage Networks in Austin, Texas, and a member of HP's Encompass user group, says IT managers looking for options are zeroing in on these new virtualization and clustering technologies as they move toward utility computing.
"As different platforms evolve and become more mature, the focus on the hardware and applications is de-emphasized," he says. "You're more focused on the actual business you're trying to conduct and therefore the continuity of that business. ... You no longer care what's in the background, you just need that business service to be available."