Exec: HP blades get fault-tolerant feel

Scott Stallard, senior vice president and general manager of Hewlett-Packard's enterprise storage and servers group, spoke with Computerworldhere last week following the launch of the company's new BladeSystem c-Class hardware line. Excerpts from the interview follow:

What will HP be telling its rack-mounted server buyers? Are you going to be saying that the Blade'System line is your best option? Most customers know what they want versus us steering them in any particular way. If you're a small to medium-size business customer and you don't even have a rack, you know you need a pedestal [system], and that's why we are still selling them. If customers have a more modest data center, they are still going to buy a rack. Where we see blades coming into play is for customers like ourselves -- we're going through this big [data center] consolidation. We're tired of 30 percent utilized stuff sitting around everywhere with too many operations people -- we're just going to take it all and put it in this [blade server] footprint.

The other place you see [blades] is when companies go to build out, as an example, remote replicated sites, [where] you just need an infrastructure in a box and would like to manage it remotely.

You said you're using technologies from HP's NonStop fault- tolerant systems in the Blade-System line. Which ones? We brought over some of the interconnect technology that is part of Virtual Connect [a virtualized wiring architecture built into Blade'System]. We also took some of the technologies in hot-plug interfaces. In the NonStop environment, you can never take your system down to accommodate [changes in devices].

Do you plan to build a NonStop blade server? It's going to take a few years, but our plan is to run NonStop -- 100 percent-availability-type systems -- on these blades that are specially made but still in the c-Class chassis and that have either dual- or tri-modular redundant capabilities. It completely changes the economics equation for NonStop computing. Things that used to be only the domain of banks or stock exchanges -- those things will not be one-off proprietary systems but instead will be environments of these industry-standard blocks that we deploy dynamically.