NEC: Entry-level Fault-Tolerant Servers Reduce Need For Clustering

The company on Monday unveiled the single-socket version of its Express5800/320Ma fault-tolerant server line, its first such server to be available starting at less than $12,000, said Mike Mitsch, general manager of the Irving, Texas-based company.

NEC's fault-tolerant servers are based on the company's lock-step technology under which two processors simultaneously work on the same operation so the failure of one processor does not impact the operation. The servers also have redundant power supplies, fans, and other components to ensure "five nines" or 99.999 percent availability, Mitsch said.

The new server actually has two sockets, each of which is populated with a 1-way SMP Intel Xeon 3.2GHz processor to form two logical servers inside the physical server. Other models in the family can be configured as two dual-socket 3.2GHz Xeon logical servers or as a two 2-way, 2.8GHz dual-core Xeon-based server.

Fault-tolerant servers have always been an attractive alternative to clustered servers, which require much time to deploy and maintain and require specialized IT talent, Mitsch said.

However, with the new entry-level model, it is possible to deploy fault-tolerant servers at a price lower than clusters, he said.

"We now offer our base configuration at a cost lower than the cost of clustering," he said. "So out of the box, it's less expensive than an equivalent Intel-based clustering technology. So VARs can offer fault-tolerant technology at the same price as clustering technology without the cost of development, support and operational complexity, and in many instances achieve a five-nines level of availability."

Benjamin Woo, vice president of sales and marketing at ASI System Integration, a New York-based solution provider and NEC partner, said that the biggest challenge in high-availability is the clustering software.

"If a customer wants fault-tolerance, they look at clusters," Woo said. "But they don't want to customize the cluster every time. Smaller customers don't have the expertise."

However, since not every application needs to run on a fault-tolerant server, even smaller customers can pick-and-choose which of their applications are critical enough to move to a fault-tolerant server at the new price point, Woo said.

"Even smaller customers can us a fault-tolerant server for applications like Exchange," he said. "If they do that, why spend money on a high-end server? This is cheaper than clustering. With clustering, they would need two servers, cluster-aware software with extra licenses and people with clustering expertise."

However, lower price alone will not expand the market for fault-tolerant servers, said Sam Bishop, business development director at All Computer Solutions, a Portland, Maine-based solution provider and partner to NEC and Stratus Technologies, another developer of fault-tolerant servers.

The lower price is absolutely a great idea for customers, Bishop said. However, the biggest challenge is the same as it was years ago: There are not enough companies and IT people aware of the advantages of the technology.

"You can't assume people will understand the technology," Bishop said. "Customers might not even know the technology exists. It's not like when a car manufacturer comes out with a new car. Everybody knows cars. With fault-tolerant servers, it's not about a price point or a snazzy new product. It's about getting people to know the technology. You've got to get the word out."

Woo said that customer understanding of fault-tolerant technology is actually quite good because of the widespread understanding of business continuity and disaster recovery.

"If you question why you need fault tolerance, it's like questioning why you need RAID," he said. "If you are a small or a medium business, the cost of disaster recovery is prohibitive. But without fault tolerance, if a server goes down, the company can go down."

Fault-tolerant technology offers many advantages over clustering, Woo said. For instance, in a clustered environment, users are constantly patching multiple servers, and still cannot be sure a patch won't break the cluster, he said. And customers who build clusters based on Dell servers, which he said is known to ship multiple servers of the same model built with different components, run the risk of having to do different patches for the servers in the cluster.

NEC's fault-tolerant servers vary slightly in terms of specifications from those offered by Stratus, even though the two cooperate in terms of product and technology development, Mitsch said.

"We have a long-term relation on lock-step technology," he said. "Stratus is focused on specific verticals. We also have our vertical focus. It's a case of coopetition. It's important for the technology to be available from multiple sources."

NEC has an equity stake in Stratus, and in November 2005 made its most recent investment, $9 million, in the company.

Orders for the new fault-tolerant servers are being taken this week, with delivery slated within standard lead times, Mitsch said. They are only available through the channel, he said.