The futility of utility computing

Back in the 1980s, the folks at Digital Equipment had a problem. While their VAX systems were selling like hotcakes, the systems still couldn't deliver the kind of scale found on mainframes.

Digital, which had no plans to build big iron, set out to find a way to use existing equipment to increase capacity--a search that led to the development of the VAX cluster and the concept of horizontal scaling.

Rather than roll in bigger and bigger boxes, Digital discovered a way to balance compute loads across multiple machines. What's more, the systems behaved as a single system image so IT administrators could manage the cluster as a distinct entity. This masked the underlying complexity, lowered cost, and simplified day-to-day activities.

This history review provides a good analog for today's computing utility vision. HP, IBM, Microsoft and Sun Microsystems all have major initiatives underway to ease complexity with their own computing utilities. (It is ironic that these same companies contributed greatly toward building this complexity, but that's another story.) The fact is that in spite of good intentions and armies of engineers, none of them will achieve anywhere near the success of the venerable VAX cluster.

Remember that Digital was a vertically integrated company that controlled all the parts, including the hardware, the operating systems, the interconnects and the database. Getting these pieces to work together in harmony was difficult enough when a single company is in full control. It is next to impossible when there are literally thousands of possible combinations.

How will the system vendors handle this challenge? Open APIs? Industry standards? Joint development? So far, they?ve been relatively quiet on this point. I doubt that technology leaders like EMC, Cisco Systems and Oracle want to develop, test and support multiple versions of their products that support each computing utility model.

Unlike the simple applications of 20 years ago, today's N-tier applications cross multiple computing environments and are susceptible to constant performance and operational issues. So you can imagine a situation where a company?s back-end MySQL database running on a Solaris N1 computing utility may be load balancing and using its resources efficiently while the IIS-based Web server tier is completely choked.

Perhaps system manufacturers believe that their utility computing systems will persuade customers to migrate off other platforms in favor of their own? Don?t bet on it. It?s more likely that enterprise shops will have a mix of multiple revisions of each vendor?s computing utility software. This ugly situation may be even worse than today?s mess.

Another thing to consider here is that the VAX cluster lived in a much simpler time when central IT controlled computing resources, communications and a limited number of back-office applications. Heck, some of them even still wore lab coats.

CIOs can't count on whiz-bang computing utilities as a panacea to today?s technology complexity and high costs.

By contrast, today's IT organization consists of a conglomeration of system, network, application, operations and security groups; each of which is armed with its own set of management tools. As part of their responsibilities, these groups must touch each server constantly. Add more servers without adding staff and things are bound to unwind. Solving technical problems means getting these disparate IT groups to come together and cooperate. Any IT manager will tell you that this is easier said than done.

CIOs can't count on whiz-bang computing utilities as a panacea to today?s technology complexity and high costs. While they wait for the next big thing from their system vendors, they should get busy addressing tangible organizational and operational issues.

What to do
• For starters, break down traditional walls between IT functional groups and organize IT around critical business processes, services and applications. Mandate that these blended IT teams review all policies, processes and procedures to look for inefficiencies, redundancies and points of contention. Replace these broken processes with well-defined IT governance and standards. This will go a long way toward improving communications, cutting costs, and enhancing technology performance and availability.

• IT groups should also examine their management tools portfolio and look for ways to support this business focus. Tools from the likes of BMC, Computer Associates, Smarts and Tivoli can be configured to provide a business process view. Instead of monitoring systems and networking equipment, these tools will aggregate technical information into a complete view of an application environment. To help lower cost and improve service, companies that already use these tools for system and network management should consider this expanded role.

• Finally, IT managers should explore the hundreds of new tools available for operations management. The options range from suites from BladeLogic and Opsware to point solutions for problem areas like server provisioning (Configuresoft), patch management (Citadel) and security (Tripwire). There is tremendous venture capital investment and innovation in this area.

Computing utilities appear to be the current golden goose, but their development will take years.

It's human nature to try and find the easiest possible solution to a problem. In this spirit, computing utilities appear to be the current golden goose, but their development will take years. CIOs can expect marginal advances and continued hype for the next 12 to 18 months. Over the next 18 to 36 months, advanced computing architectures will emerge but their application will be limited to bleeding-edge applications. Mainstream progression in this area won't arrive until 2006 and savvy IT executives won't wait around for the next VAX cluster. Rather, they will repair organizational and operational woes. Addressing these areas will certainly offer more benefits than vaporware or vendor rhetoric.