Stabbing Your PCs

Why not just use server-based computing, a la Citrix?

What is the advantage to blade PCs v. other solutions?

Can users of blade PCs have their own user profile?

What about cost?

We can answer the last first. HP claims that the list price, per seat, for an IT setup using blade PCs is $1399. That includes the blade PC, the infrastructure to support it (like racks and enclosures, along with the control software) and a thin client for each user who needs to get access to a blade PC.But we're getting ahead of ourselves. First, what is a blade PC?

It's a blade computer that uses a processor bought from Transmeta Corp. called the Efficeon. It's an x86-compatible processor that is claimed to be more efficient than similar processor from other vendors. The blade PC goes into a rack in the data center and gets accessed by the user from a thin client. At logon time, an allocation engine couples the user's thin client to the appropriate blade PC, and associates that user's user profile with the blade PC in question using the resources of Active Directory. The blade PCs use embedded Windows XP as their operating system, and because the thin clients are Windows-based, the user gets the full Windows user experience.

Tad Bodeman, director of thin client and CCI (Concolidated Client Interface) solutions for HP, explains why the blade PC is prefereable to server-based computing in some instances. "Today, in our server-based solutions," he says, 80 to 90 per cent of the clients are desktop PCs. You're trying to reduce cost of delivery and supporting applications to users. But the majority of applications are legacy custom written line-of-business apps, and they were written for PCs. That means they were written under design parameters that assume full use of all (host) resources."

That would be fine, he continues, but in server-based computing, the application won't have full use of all the host resources, because there will be other sessions running on the server for other users, and they will want to use some of those resources as well. So there is some requirement for rewriting apps to port to server-based computing, or some apps just won't work at all. "You just can't do it," he summarizes. That's why so many clients in server-based computing are PCs. The local PC can run the applications that don't run well on the server.

Further, when you load up a server-based computing implementation with lots of users, things slow down. But, claims Bodeman, because each user in the CCI setup with a blade PC and a thin client, everybody gets his/her own blade PC, so there's no slowdown when many users are logged on.You can see, though, from this discussion, that you need more than just the thin clients and blade PCs for all this to work. You need some server, somewhere, to handle the resource allocation--maybe a blade server. You need storage, because user profiles have to be stored somewhere, as do applications, data, etc. But you don't necessarily need one blade PC for each user. Studies show that perhaps eight out of ten people are running their PCs at any given time, so you can use fewer blades than people. "You can have a nice policy-driven rule for adding blades," says Bodeman, "so that when you add, say, up to nine people, you add a blade, but you don't with the tenth person." This can save more money.

Bodeman says HP thinks that CCI and the blade PC will be complementary to the server-based computing that, so far, he says, has only limited acceptance. He thinks the two together give the benefits of the PC experience, along with the benefits of centralized management and lower TCO.

Could be. Time will tell, as it always does.

David Gabel has been testing and writing about computers for more than 25 years.