Ever find yourself with one too many storage servers when you actually needed another application server?

A company called Egenera thinks it has a solution to that dilemma: A sort of existential blade server. It's a server that's neither this nor that but could be almost anything at hardly more than the blink of an eye.

The concept is called the BladeFrame and it's the brainchild of Vern Brownell, the former CTO of Goldman Sachs. During his ten year stint as manager of what could arguably be called one of the world's most complex data centers, Brownell looked at what he termed the "pain points" from a customer's perspective. For some of you, facing pain may be an alarming idea. Just take a few deep breaths. After all, if you're annoyed by the things that slow your network to a crawl or cause it to go down in flames, just think about the aggravation such events inspire in the people who need to access your network.

What Egenera proposes with BladeFrame is a bank of available blade servers (using industry standard commodities from Intel, Microsoft, and Linux with I/O capabilities similar to InfiniBand) that start life with no specific identity--a Processing Area Network, as it were. There's no storage. That's where your SAN comes into play. You could have as few as two or as many as 24 blades in a rack, plus two control blades and two switch blades. You can mix and match processing blades in a single BladeFrame: Let some blades become Web servers, others can act as database and application servers, and you can still have room for a customer server with a few blades left for failover or additional capacity as needed.

That almost sounds traditional, doesn't it? Well, not really. The I/O is routed through the control blades to minimize cabling, and virtual switches and interfaces are used to replace their physical counterparts. As a virtual entity, the BladeFrame can be configured logically through software (BladeFrame runs an optimized version of Red Hat Linux). Egenera also provides a management layer that uses self-correcting diagnostics that can be programmed to respond in a variety of ways to a variety of potential problems--from pending work overload to that first whiff of acrid smoke. And because the components are anonymous until the software assigns them their identity, everything is fully redundant (including highly available, load balancing clusters).

How does it all impact TCO? In a 24-server environment, Egenera claims probable up-front savings--based on the cost of the ancillary hardware alone--in the range of $219,000. Egenera backs up its claim by tallying the equipment you'll no longer need: For a traditional 24-server environment, Egenera says you'd need 48 storage and Gigabit Ethernet ports, 24 terminal ports, and 48 host bus adapters (HBAs) and Gigabit Ethernet NICs. Translate that configuration to a 24-server BladeFrame and the additional hardware requirements drop sharply to just 2 storage, GigE, and terminal ports. The HBAs and NICs are already included.

It gets even better according to Egenera. Choosing a BladeFrame over a comparable Sun Fire 280R would put $962,000 back in your budget, while saving annual recurring costs (admin, floor space, and so on) of about $158,000.

Of course, these are Egenera's numbers and undoubtedly there's some optimism included in their tallies. Since vendors tend to use the absolute worst and the very best scenarios when making a case for their products, it falls on you to run the numbers. Still, if you could approach upper management with a plan that might save them almost $1.8M over the next five years, they might do something silly like cut your budget by only 30 percent instead of 50 percent. Who knows, your network customers might even show their gratitude with increased business.

Would your company consider a BladeFrame over a Sun Fire 280R? Why? Tell us in our TalkBack forum, or send a note to Bill.