Guild Companies

Newsletters

Subscriptions

Media Kit

About Us

Contact

Search

Home

Windows & Linux Edition
Volume 2, Number 18 -- May 7, 2003

Palmisano Outlines On Demand for Shareholders, IBM Rolls Out Products

by Timothy Prickett Morgan

IBM chairman and CEO Sam Palmisano has seen the future of the IT industry, and it is black and blue. The black part is not the abyss, but rather profits; the blue part is, as you might expect from Palmisano, Big Blue itself and its central role in what IBM is calling e-business On Demand. You and I might call it distributed computing as originally intended but never delivered. Of course, when addressing shareholders at its annual meeting in Kansas City several days ago, Palmisano didn't come out and say that. He likes On Demand better.

If there is one thing we don't need in the IT industry, it's another buzzword that conveys some meaning, but not quite enough. But we're stuck with "on demand" just like we were stuck with "e-business" in the late 1990s. IBM is not just blowing smoke here, so don't get me wrong when I criticize the term "on demand." I know lots of companies that are trying to figure out who can build what amounts to an information utility, because they are sick to death of buying servers and software from any vendor, they are sick of having to buy three times the normal capacity, to cope with peaks, and then have it sit there idle most of the time, and they are sick of having IT be a fixed cost, rather than a variable one, that rises and falls with use (like electricity, water, telecom, and other utilities).

IBM's On Demand vision is obviously a self-serving vision, one that plays to IBM's largess and its hubris, and big business is probably going to eat it up. But if IBM thinks there won't be a zillion information utilities established, and cut-throat competition among them, it is loco in the coco. One only need look at the commercialization of the Internet, and the proliferation of some 10,000 service providers in the United States alone in the 1990s, to see what is going to happen. Many predicted rapid consolidation among ISPs, especially as AT&T, AOL, MSN, and others joined in. But guess what? It didn't happen, because big e-mail and Web service providers don't cover every niche, and they don't please every potential customer. Information utilities, if they are created and thrive, will be no different.

Big companies will build their own utilities or have IBM Global Services or Electronic Data Systems do it, but most companies will probably just find a service provider to augment their internal machines and their applications. A large part of what On Demand is about is fulfilling the promise of application service providers, or ASPs. (Remember that acronym?) We've heard all of this before. But now the distributed computing and application service providing vision has been jacked up one level to Java and Microsoft .NET Web services and implemented (in theory) with grid computing, resource virtualization, and autonomic-systems-management technologies. None of this is really available today, but pieces of it are. You can do some On Demand stuff today with IBM gear, and, by the way, Hewlett-Packard and Sun Microsystems have some pretty cool stuff, by virtue of their Utility Data Center and N1 efforts. Sun ruled the ISP business for years, and that is what fueled its growth, and if utilities proliferate and Sun can get its N1 products out the door, it has every shot at giving IBM a run for its money. To HP's credit, UDC is a real product today, but HP doesn't seem to know how to sell it. No one will be accusing IBM of that, as we have seen from recent TV ads, which are as funny as the old e-business ads were.

Palmisano made the case for IBM's triumph in the utility era, and the prospects for the IT industry, to shareholders:

There has been speculation, in recent times, about the IT industry and its future. Some have wondered if its best days are over. Some have speculated that a trillion-dollar industry simply can't keep growing. Let me reassure you: This industry remains attractive, important--and it will remain a growth industry. In fact, we believe that, long-term, it will continue to grow faster than GDP.

However, it's important for all of us to understand that future growth will not be stimulated by technology alone. This is a historic shift for an industry that has always been fueled by a constant stream of new technologies, products and gizmos. But growth will not be fueled that way in the future. Why? Three reasons.

One: IT is a $1.3 trillion industry. It's wishful thinking that a single technology will drive growth for the whole industry. IT has gotten too big, and too interwoven into the fabric of business, for that to happen.

Two: Although technical innovation is, and will always be, a driving force of this company and the IT industry, customers don't value technology for technology's sake; they value the application of technology to solve serious business problems.

Three: Customers no longer think about computing as a collection of piece parts. Today, computing is viewed holistically, as a technical architecture, an end-to-end system, a computing model.

The computing model doesn't change very often, but it's changing now. In the past, customers used technology to automate stand-alone operations like payroll and inventory control--the back office. Or procurement, manufacturing, the Web-based store. Customers have derived great benefits from applying technology to those business processes. But today customers want to use technology to pull those stand-alone operations into a unified whole, to boost productivity and deliver tangible returns. They want to leverage network technology to help them build a business that can respond dynamically to whatever the world throws at them. And they know that they can't build such a company on top of stand-alone computing systems and piece-part technologies.

Heaven knows, we sure have been trying to build flexible companies on top of rigid and isolated computing strategies that promised openness and portability but never quite delivered what we expected. On this, we would probably all agree. In the past, the systems that we set up with our computers embodied the processes that the people who are behind a business wanted to use to conduct that business. Encoding those processes in any language, on any system, was a severe commitment: If you needed to change something, it took effort and money, and, in some cases, an advanced technology was not available on your platform, so you had to buy another one and integrate them. In the future, IBM, Sun, HP, Microsoft, and others talk about how applications will be able to roam around and run on any platform, cluster, or whatever that is available to them. I like this idea. I love it. But until some pretty hefty standards emerge, every vendor's utility is going to be incompatible, just like the electric and cell-phone systems in the United States and Europe. It's the nature of companies to stay as proprietary as possible for as long as possible, because that is where the easy profits are.

That said, a utility-style of computing with variable costs is a lot better for many companies--for at least some of their workloads--than building personal data centers, so even if the utilities that IBM, HP, Sun, and others are building are incompatible, there are benefits that make utility computing something worth keeping abreast of.

Today, at an analyst conference hosted in the Catskills, in New York, IBM rolled out a bunch of new On Demand offerings, expanding the basic permanent and temporary processor upgrades on demand that it offers in its zSeries mainframes and iSeries midrange boxes. IBM has capacity on demand (COD) in its pSeries line, but this is only for permanent upgrades, not temporary ones. This lack of temporary usage of processors in the pSeries line--kind of the basic starting point for the On Demand effort, as far as servers go--will apparently be rectified with the impending pSeries announcements, which will boost the performance of the pSeries line with a new 1.7 or 1.8 GHz version of the Power4+ processor as well.

Specifically, IBM is rolling out standby COD in the BladeCenter chassis, and will allow customers to buy a machine with seven active blades and seven inactive HS20 blades, allowing them to activate the extra blades as they need to. (The HS20 blades are a two-way server based on Intel's "Prestonia" Pentium 4 Xeon DP processors.) Customers are being given a six-month window to activate those additional seven blade servers. IBM is not allowing customers to use the blades in a true On Demand fashion--meaning they can turn them off when they are not using them. Products have not, it seems, caught up to strategy in the On Demand world IBM is trying to build. None of this undermines the usefulness of the COD offering on the BladeCenter, but this is not On Demand computing as described by Palmisano. IBM is also offering COD on its Shark storage arrays, jamming an extra 6.9 TB of capacity into an array, which can be activated as needed. These offerings will be available later this year; exactly when is unclear.

On the pricing front of the On Demand strategy, IBM hardware and software technologies will soon be available under a new kind of contract called the Open Infrastructure Offering, which allows customers to buy some or all of their infrastructure for a single fixed monthly cost. (Hey, wait a minute. Wasn't On Demand all about turning fixed costs into variable costs?) Under the Open Infrastructure Offering, customers can pay a fixed amount for IT gear, but they can upgrade that technology as needed to new hardware or services. Exactly what technologies are available under an Open Infrastructure Offering contract, and what that contract costs or what elements can be upgraded (and to what and when), were not divulged at the announcement, but an IBM spokesperson says it covers all IBM products and services and is available only on a special-bid basis.

IBM's On Demand strategy depends on the ability to virtualize components that make up systems and networks. IBM's Tivoli systems-management-software unit has been working on a Web Server Provisioning package for its BladeCenters that makes use of the IBM Director and WebSphere middleware, as well as Tivoli Storage Manager and monitoring software, to do rapid deployment of Web servers. IBM is also working on a User Provisioning offering that sits in front of network resources to manage the way that users gain access to network resources and applications. The User Provisioning offering is based on Tivoli Identity Manager, Tivoli Access Manager, and Director Integrator software, and runs on IBM's pSeries Unix boxes and on its BladeCenter blade servers.

IBM also announced that its WebSphere middleware has been grid-enabled, something it promised when it backed the Open Services Grid Architecture (OSGA) Web services initiative last year. The IBM Server Allocation for WebSphere Application Server product was a joint development project by IBM Systems Group, IBM Software Group, and IBM Research. This product will be available later in the second quarter and will be based on WebSphere Enterprise Edition. It will cost around $100,000. IBM says that Tivoli systems management programs will be grid-enabled sometime in the second half of 2003. At the analyst event, IBM also previewed a set of storage virtualization technologies formerly called "StorageTank" and now known as the TotalStorage Virtualization family. The SAN Volume Controller, code-named "LoadStone," is a two-way xSeries server running an embedded version of Linux and IBM's StorageTank virtualization software; an entry configuration of this machine, which will be available in July, costs $75,000. The SAN Integration Server portion of Storage Tank is a complete SAN setup including the SAN Volume Controller, up to 83 TB of storage (using IBM's FAStT600 arrays), redundant Fiber Channel switches, and various routers and hubs. The SAN Integration Server will be available in August.

Perhaps most significant, IBM will eventually deliver a SAN File System running on a bunch of Linux-based xSeries machines that runs other StorageTank components and delivers a heterogeneous file system with file-level locking--meaning it can support all different kinds of operating systems (Unix, Windows, Linux, mainframes) and that the files can be shared by these incompatible operating systems simultaneously. This latter item is sorely needed and has been one of the limiting factors in distributed computing. If disparate machines can't share the same data, then distributed computing just means islands of automation that are linked together with fast networks that can't really share the computing responsibilities.