Although commentators marvel that Moore’s Law has remained in force for four decades, rapidly doubling computer power every 18 months or so, few of them dwell on the fact that our ability to consume this exponential gift of computer power easily outstrips the ability of the computer industry to deliver it. And yet this is exactly what has happened in the 40 years that Moore’s Law has held sway.

The perpetual increase in computer power bestows enormous benefits. The speed of transaction processing multiplied by about 10 every six years, enabling us to move from a single mainframe computer to client/server systems, then to multi-tier server systems and then to systems deployed across the Internet. The data that we put into databases expanded even more dramatically, by about 1000 every six years. In 1991 we managed megabytes then by 1997 it was gigabytes and by 2003 it was terabytes. Soon it will be petabytes.

This is one reason why we consume the abundance of computer power so easily. Applications that were leading edge 6 years ago, because of they were power hungry, are accommodated easily and cheaply now and will be considered almost trivial in another six years. Anyone who wants them will have them. But this alone doesn’t explain the dramatic growth of corporate data centers across the world.

Before the advent of client/server data centers had very few computers, perhaps a couple of mainframes and a handful of minicomputers. The numbers started to grow with client/server deployments, accelerating quickly with the advent of the Internet, so that 10s of computers grew to 100s and then to 1000s and in some cases headed towards 10,000 and beyond.

The reason the data centers grew so swiftly is twofold. On the one hand, the number of applications we deployed expanded, so organizations that once ran maybe a few hundred applications now run several thousand. On the other hand computer hardware grew less expensive every year, so cheap servers proliferated and we made no effort to use them efficiently.

The Cobblers Children

As the data centers have grown, so have their costs, including the cost of computer room space and of electricity to run the computers and to keep them cool. And yet the biggest cost of all is the labor cost – not just the labor costs of running the data centers, but costs of the labor needed to support the data center – the help desk, the desktop administrators, the networking specialists, the DBAs, the security technicians, the storage management staff and the rest of the army of support that a data center now requires.

Nowadays labor costs make up about two thirds of the cost of running a data center (we should have the associated graphic from Al’s presentation here). While IT departments engage in server consolidation projects to reduce the number of data center servers and computer manufacturers work on reducing power consumption and cooling costs, another area of focus that is desperately needed is in better automating the management of the data center – not to reduce the number of staff involved in operations, but to try to keep the numbers from rising, because right now the number are growing by about 10 percent per annum.

SOA – The Tipping Point

It should be no surprise that the number of IT support staff is mushrooming. Most of the management software that is currently deployed was written to manage data centers with relatively small networks of computers. It was never intended to work across hundreds or thousands of computers. And at the heart of system management is software that does little more than report when a hardware or software component fails. Beyond logging events, it does very little to help diagnose what the underlying problem is and consequently it can do nothing whatsoever to fix the problem automatically.

Consequently business applications have tended to be implemented in siloes, almost as stand-alone systems – so that managing the system while in operation, backing it up and recovering it if it fails are handled on an application by application basis. When organizations run so many applications, this approach is already questionable as it leads to over-provisioning of hardware and poor standardization, which in turn necessitates greater manual activity. It can hamper the use of virtualization to consolidate servers and reduce the server population.

With the advent of Service Oriented Architecture (SOA) a tipping point is reached. SOA enables the reuse of software components and encourages them to be linked in an end-to-end manner, integrating siloes one with another. This software architecture is proving both popular and successful, because it enables faster building, deployment and change to software and because it aligns business applications with business processes. It has also become an inevitable direction for corporate software because all the major software vendors are now adopting the Web Services standards that are fundamental to SOA and are in the process of delivering products that are SOA ready. But what will IT Departments use to manage the SOA business applications that are already being built and that will eventually span the corporation? Not the traditional management software of the past.

The IT Management Crisis in Summary

The number of applications that the corporate IT operation has to support has been increasing rapidly and it continues to increase. Mobile technology is now being widely deployed and soon VoIP will be added into the mix, and video, and stream processing and also RFID applications and embedded chips. The signs are that we will continue to outstrip Moore’s Law for the foreseeable. At the same time, SOA is creating business applications which simply cannot be supported by the IT management software of the past and the rise in the number of IT support staff will, at some point become unsustainable. There is an IT Management Crisis building and dealing with it demands new thinking.