Commentary: Cutting to the chase on blade servers

By Galen Schreck, Analyst

The server industry is in a blade frenzy, and vendors have been quick to connect blades with the utility computing hype. But will blades really pay off for customers who don't need high-density racks of servers? Yes--eventually.

The rise of modular systems and pay-as-you-grow pricing has focused server vendors on building the ultimate box for utility computing. But many buyers wonder if the modular computing hype is just a temporary reaction to a tight spending environment. They ask:

Are blade servers the future of computing or just another fad?
While trendy now, blade servers are here to stay--but their value is only partly tied up in their shape and size. In fact, it's their operational efficiency that will keep them around. Already, smart management software can rebuild the contents of a failed blade when a replacement is inserted in the chassis. And tomorrow's blades will go one step further, ditching their internal drives entirely in favor of a storage area network.

		Related story Utility computing reaches HP PCs, printers The company has been working feverishly to add specific products and services to its utility computing idea, an admittedly confusing strategy.

Are blade servers right for all types of application loads?
Almost--most applications scale well on today's blades, although some databases, for example, may benefit from servers with more than the four central processing units found on today's biggest blades. The net result? Blades work well for scaling out in the Web and application tier, but companies that can't use clustered databases like Oracle 9i RAC will still need plenty of 16- and 32-way systems.

Don't blades cost more than "regular" servers?
It depends. When fully populated, most blade frames are very close in price to a comparable number of rack-optimized servers. The catch? Buyers pay for every slot, whether they use it or not. For example, each slot in IBM's 14-slot blade chassis costs about $350, and the blades themselves can be up to 20 percent more expensive than rackable servers. On the other hand, products like Egenera's BladeFrame can substantially reduce network port costs by consolidating all I/O for the frame over 2-gigabit Ethernet and two Fibre Channel ports.

Do blade servers really simplify data center operations?
Yes. Today's blade servers have highly integrated management capabilities that aren't yet available for stand-alone servers. For example, when a dead blade is replaced with a new one, the chassis can alert management software to the new gear, kicking off a series of automated provisioning events. With automatic provisioning and no racking or cabling to mess with, blades can be up and running quickly.

When are rackable "pizza box" servers a better buy?
When floor space isn't at a premium, density doesn't matter. And it doesn't make sense to pack servers into a tight space if massive amounts of cooling aren't available. Finally, small information technology shops--and small offices of large companies--probably don't need the sophisticated management capabilities blade servers enable.

Aren't today's blade servers highly proprietary?
You bet. Today's blades and management tools--from a single vendor--are designed to work tightly together. Better industry standards for blades will simplify this by 2005, but the truth is that blade hardware and software are still in flux. Just one year ago, vendors were focused on superdense systems with lower CPU speeds--while now, it's about faster Intel Xeon processors and more of them for power-hungry applications. The bottom line: Management tools for blades won't be standardized for another two years.

© 2003, Forrester Research, Inc. All rights reserved. Information is based on best available resources. Opinions reflect judgment at the time and are subject to change.