Transforming the datacenter

THE LOOK OF COMPUTER datacenters will begin to change as a wave of server-blade products begin hitting the market during the next several months.

Following the lead of startups such as RLX and Racemi, server manufacturers including Compaq, Hewlett-Packard, IBM, and Sun are each in the process of rearchitecting their own server offerings into the dense, rackable blade style.

The potential for server blades lies in their capability to not only give customers managing large server farms a way to reduce the amount of space needed to accommodate vast server resources, but also to offer the industry a more logical and efficient way to add computing power as it grows.

Experts agree that server blades will give companies an edge by consuming less power and taking up less space than traditional servers. Advanced server blade software-management tools will also rapidly enable virtual-resource pooling in server blades assigned to application-and database-level responsibilities, paving the way for a whole new way of thinking about the datacenter.

“Blades are a part of a complete hardware and software infrastructure that incorporates load balancing and fail-over, resource redeployment, and a more virtualized computer platform,” says Gordon Haff, a senior analyst at Illuminata in Nashua, N.H.

Haff says that beyond the initial space and power savings prompted by server blades, blade architecture sets the stage for the intersection of all network and datacenter components, which could attach to a single, easily managed chassis or backplane.

Blazing that trail, San Jose, Calif.-based Compaq this week rolled out its first server blades, the ProLiant BL Line. Dense enough to pack 280 server blades into an industry-standard server rack, Compaq’s ProLiant BL blades are part of the company’s AIS (Adaptive Infrastructure Strategy), according to Compaq officials.

AIS calls for the addition of advanced fail-over and clustering technology, remote system management, self-healing technology for server hardware, and system automation to blade-server products. Compaq plans to eventually offer a wide range of server blades for not only front-end applications, but also for application-layer environments and back-end database deployments.

Within weeks, Palo Alto, Calif.-based Hewlett-Packard will begin to expand the blade concept beyond that of mere servers with an expected line of blade hardware that will include storage, switching, and network management blades.

IBM, in Armonk, N.Y., is also developing blades from the hardware and software angles, according to Brendan Paget, worldwide marketing manager for IBM’s xSeries server group. As with Compaq, IBM plans to extend server-blade architecture all the way to the database. The company is well into the development of server blades that house “a totally self-contained computer” on a single blade, Paget says.

Early this year, Big Blue will also begin to offer what Paget calls “virtual blades,” in which a server-blade system could be partitioned internally to not only have separate, hot-swappable hardware components, but virtual software partitions that could be reallocated for fail-over on the fly.

Sun has already incorporated blade architectures within its high-end StarCat server, which has modular, interchangeable processor blades that automatically format themselves to the environment and application as they are installed.

But early adopters of server blades should be aware of the absence of a common, cross-vendor standard for the backplane interconnect of first-generation blades.

John Enck, a senior research director at Gartner in Stamford, Conn., notes that server blades may not realize their greatest promise until companies can agree on a standard for the backplanes of server blades, allowing users to swap out one company’s server blade with another.

“There is no standard for what the blade interface and what the chassis is. Every vendor is different. So there is going to be a big war over standards in the next 18 months,” says Enck, adding that HP is one vendor trying to lead the cause and establish a standard starting point with the PCI standard.

HP’s “Blade Program” preaches adoption of the CompactPCI and NEBS (Network Equipment Building Standard) standards as the basis of all server-blade architectures. Compliance to these two standards would create fully interoperable server blades from multiple vendors that can be interchanged on the fly.

“We are basically unleashing the creativity of the industry on the standards-based approach,” says Brian Cox, worldwide product manager for entry-level servers at HP. But as of now, HP still stands alone in its determination to build blades to the CompactPCI and NEBS standards.

“In general, anything people buy at this point in time in terms of blades really is a vendor-specific solution,” Illuminata’s Haff explains. “It’s a caveat in the sense that you are buying and adhering to specific types of interconnects.

“Not to say that you should not buy blades today if you have a business need for them, because they can solve a number of problems in terms of power and density,” he adds. “But you should go with your eyes open and assume that if you buy an HP solution today, it will only plug-and-play with other HP blades. That could all change, but we are not there today.”