HP Launches Blade PC, Other Adaptive Enterprise Gizmos

by Timothy Prickett Morgan

Every now and then, an idea so obvious comes along that you wonder why you didn't think of it first. Hewlett-Packard last week unveiled what it is calling a blade PC, along with a bunch of other virtualization offerings that have been pushed under the Adaptive Enterprise umbrella. The blade PC concept is interesting, and it could catch on in a big way.

HP has been trying extra hard to prove that Adaptive Enterprise is a real thing, and its executives have spent a lot of time in recent months trying to convince customers, competitors, Wall Street, and maybe even itself that the Adaptive Enterprise initiative is more than a lot of marketing speak. The simple fact is that there is no arguing that IT infrastructure is expensive to acquire and maintain and it is often used in a very inefficient manner. Adaptive Enterprise is an approach to trying to fix that which includes virtualization technologies and business process engineering, among many other things. IT is hard to describe and manage, so it stands to reason that any set of products and processes that are created to control it would be difficult to describe in a sound bite. Eventually, vendors may stop trying so hard to convince us of their vision and actually get more virtualized, on demand, utility products out the door.

The blade PC is an actual product, or at least it will be when it becomes generally available in March 2004. It is not a new concept, but being the volume leader in desktops and servers will undoubtedly help HP create a blade PC market and dominate it if customers are interested in the concept. The blade PC is a twist on the now mundane concept of the network computer that links into network-enabled applications that reside on central servers. The first time some of us heard of the blade PC variant on network computing was from ClearCube, an Austin, Texas, which was founded in 1997 and which has been pushing a rack of what are blades based on uniprocessor Pentium 4 and two-way Xeon DP blades with fast graphics cards. Up to 112 blades can go into a rack. End users connect to the blades from their desktops using a set of ports that link to the rack with standard Ethernet. The blades are a shared resource, with one user per blade, and they can run Windows 2000, Windows XP, or Linux. ClearCube says that it can cut the cost of putting a "desktop" environment in front of users by 40 percent by using shared blades and desktop ports instead of real PCs.

HP is touting similar savings with the blade PC, which is being marketed under the unwieldy name of the consolidated client infrastructure (CCI) solution. HP has revamped its initial "QuickBlade" ProLiant BL e-Class servers, which could put 20 Pentium III blades (running at 700 MHz to 1 GHz) in a single chassis and 280 blades in a standard rack, with a new blade that is based on the Transmeta Efficeon processor running at 1.1 GHz. These blades, which are called the bx1000s and which have a list price of $799, are equipped with Windows XP Pro, which includes Active Directory and roaming profile features. End users log onto the blade PC through an HP t5000 thin client, which runs either Windows CE or Windows XP Embedded Edition. The blades store user Windows XP profiles and data on the blades and their associated NAS or SAN arrays, but users have the look and feel of a desktop Windows XP workstation when they sit down and work. The thin client links to the blade PC through the RDP protocol that Microsoft created for Windows, and it can link over dialup, wireless LAN, or wired LAN links to the blades.

According to Tad Bodeman, director of the CCI and thin client products at HP, the blade PC has some attractive features compared to maintaining real PCs. First of all, everything is centrally located and backed up. And the very nature of shared computing means that if a blade dies, a customer can just reboot their RDP session with their own access privileges, screens, applications, and files, in a minute or so on a new blade in the rack. There aren't the four or five different levels of service calls associated with a crash, which is what happens with a real PC. Bodeman says that the consultants that HP spoke to figure that a PC costs $4,000 to $8,000 a year over the course of three years to maintain and keep current with bug patches and such. By the most lenient estimates, a machine costs $2,600 over four years. With the HP blade PC approach, the company reckons it can knock that maintenance cost down to $1,300 a year over the course of four years. Perhaps more interesting is that HP is claiming that it can deliver a break-even return on investment in less than 12 months with the blade PC approach, including the cost of thin clients, blades, racks, storage, management software, operating systems, and such.

At list prices, these components cost about $1,500 a seat, but Bodeman figures that on the street the whole shebang will sell through the HP reseller channel or through HP direct at under $1,000 per user. This is about what a decent PC costs these days, without a good monitor. It will come as no surprise to anyone reading this, then, that HP has only previewed the concept to 100 customers and it already has a pipeline for over 1 million blade PC seats. Mark Hudson, vice president of marketing for HP's Enterprise Storage and Servers unit, says the company has not gone crazy with the thin client, network computing concept like Oracle CEO Larry Ellison has done a few times, but has nonetheless come to the conclusion that a centrally administered, blade PC approach could be appropriate for as much as 60 percent of the general purpose PC desktop market if the idea catches on. Transmeta must be beside itself with joy at hearing such numbers.

Hudson and other HP executives last week will also be banging the Adaptive Enterprise drum, with a particular bend on virtualization technologies, as part of the company's announcements. Virtualization, to HP, means a number of different things, according to Hudson. For most customers--more than 90 percent of them--virtualization currently means and will probably always mean virtualizing the capacity of a server, a storage array, or a network device so it can be carved up and more efficiently used by end users and their applications. This is serious stuff to HP, even though it is partnering to deliver a lot of its virtualization capabilities. Hudson cites Gartner analyst Tom Bittman, who said in a recent report that by 2008, enterprises that do not use virtualization technologies on Intel-based systems will spend 25 percent more money for hardware, software, labor, and floor space than those that do. Gartner reckons that the cost spread is 15 percent RISC/Unix platforms (which already have logical partitioning and decent workload managers).

To that end, HP announced a slew of enhancements to its virtualization tools for its many enterprise products:

• Pay per use capacity-on-demand server pricing, which has been available on the HP 9000 and ProLiant servers and StorageWorks storage arrays is now expanded to the Integrity line of Itanium servers, which support HP-UX, Windows, and Linux.
• Capacity on demand, which provides extra servers for standby use, has been extended to the ProLiant BL line of blade servers. HP does not offer a standard configuration for COD, but negotiates how many extra processors to put in a regular server or how many extra blades to put in a chassis with customers on a case-by-case basis. HP charges a modest fee for this capacity, but knocks the fee off when the servers are activated.
• HP's MC ServiceGuard high availability clustering software is now available for Linux running on the Integrity machines, too. It has long since been available on HP-UX servers using PA-RISC processors, and more recently was ported to the Itanium line running HP-UX. Extensions HP has created to ServiceGuard for SAP's ERP suite have also been ported to the Linux environment on the Integrity servers. ServiceGuard has also been extended for Oracle's Real Application Cluster (RAC) parallel clustered database environment in such a way that two datacenters that are 100 kilometers or fewer apart can be clustered to share work and provide fault tolerance (through geographic distribution) for the cluster.
• The HP-UX workload manager, which uses policies set by administrators to allocate resources to applications running on HP's Unix servers, can now allocate and de-allocate capacity on demand resources on an HP 9000 or Integrity server and move those resources in and out of nPar hard partitions on the machines. Hudson says that in the second half of 2004, the workload manager concept will be extended so it can wrap around other environments (Windows and Linux are obviously what HP wants to control) to provide a global, cross-network workload management tool. Windows has primitive workload management capabilities, and Linux is not much better. So this presents a challenge for HP, which controls neither of those environments.
• The HP-UX environment's Quick Start rapid deployment tool has been extended to deploy BEA Systems' WebLogic Server and Oracle's 9i database. This technology can cut deployment times in half
• On the utility front, HP is offering a messaging on demand service for Microsoft's Exchange email and groupware server in a tiered format with four different price bands with four different service levels at a set price per mailbox. HP is also launching an automated service usage service for IT organizations that want to track which users and departments consume what IT resources and charge them back for that usage.

Sponsored By UNISYS/MICROSOFT

Join the escape from UNIX. The Windows Data Center is here. The revolution has begun. The Windows Data Center: The low-cost, high-performance answer to complexity. Decisions about your data center never have to be made by default again. Introducing the proven alternative to UNIX: The Windows® Data Center. Featuring the Unisys ES7000 family of servers running new Microsoft® Windows Server 2003, it's a solution that brings revolutionary performance and low-cost standardization to the enterprise like never before. With mainframe-like integrity, the ES7000 and Windows Server 2003 optimize today's enterprise data center. It's no wonder nearly 50% of large organizations are standardizing on Windows operating systems.(1) And by focusing solely on enabling standardized, end-to-end Windows data centers, Unisys helps generate staff efficiencies, increase agility, and mitigate risks-all while achieving greater return on technology investment. So if you've had quite enough of your inflexible data center, let us help you escape. After all, nobody wants to be on the wrong side of a revolution. For More Information Call 1-800-548-3443 or visit WeHaveTheWayOut.com © 2003 Unisys Corporation and Microsoft Corporation. Unisys is a registered trademark of Unisys Corporation. Microsoft and Windows are either registered trademarks or trademarks of Microsoft Corporation in the United States and/or other countries. The names of actual companies and products mentioned herein may be the trademarks of their respective owners. (1) Unisys primary market research 1Q03.