Tread carefully - Editor's Note - Editorial

Communications News, Oct, 2003 by Ken Anderberg

Over the next two years, a great many companies will be evaluating their aging IT infrastructure and deciding which systems need upgrading and what new capabilities need to be added. Such is the inevitable result of the widespread delay of IT purchases in the past three years.

One of the choices IT and network managers will be assessing is utility computing, also known as grid computing, on-demand computing, network resource provisioning or data center virtualization. The idea is that you purchase only the processing power, bandwidth and software applications that you actually need at any given time, much like using and paying for electricity. An outsourced provider would supply the hardware and software, manage the infrastructure and bill you for only the services you use.

The advantages to end-users--at least as touted by utility computing leaders IBM, Sun, HP, EDS and Compaq--would be lower operating and management costs, and reduced complexity. The reality, however, may be somewhat different.

The lure of utility computing centers on the premise of paying only for the computing power you need and use. You basically rent server capacity and applications from an IBM or EDS, thus foregoing the need to purchase the hardware and reducing IT staffing requirements.

Application service providers (ASPs) touted the same promises but never really got off the ground. Part of the reluctance was technology related and part was the unwillingness of end-users to give up control of their important applications. The same issues confront utility computing, plus a few others.

First, the technology's not there yet. Provisioning and security protocols that will enable external utility computing are years away from implementation. With the shared server environment in the utility computing model, for example, security becomes an important concern and a technology hurdle for vendors.

Then, there's the issue of pricing. If the outsourcing model does not pan out to be worthwhile financially, IT managers will continue to keep their networks in house.

Many IT managers also might not favor losing control of their networks to an outside provider, or be limited in their choices of providers. The major vendors will provision such services using their own products, thus limiting network flexibility for many customers. Job security for those IT managers also becomes an issue if an outside provider manages the network.

There already have been some major utility computing deals announced. American Express has signed on with IBM in a seven-year, $4-billion deal for IBM to take over AmEx's technology infrastructure. EDS has signed on the 7-Eleven convenience store chain for $175 million, and HP will manage seven Nokia operations centers and 4,000 servers.

Clearly, utility computing is an option to consider. Enterprises, however, should look carefully at the cost structure, as well as issues of control and security, before venturing into this "latest, greatest" trend. A quick look at the littered landscape of failed ASP providers should be all the instruction needed for treading carefully.

Ken Anderberg

kena@comnews.com