October 12, 2007 By Larry Singer
Two primary types of consolidation are often lumped together: data centers and servers. Data center consolidation simply moves servers from multiple data centers into a central location, while server consolidation shifts the workload from many servers to fewer ones. However, each has unique challenges and different objectives.
Data center consolidation is relatively easy from a data center manager's perspective, though not from a political one. Consolidation has been part of a cyclical change between central control and distributed management of IT resources that has unfolded in government over the last 30-odd years. Part of the reason some agencies and departments have built ad hoc data centers has been to escape the high costs, poor service or political position of the central IT departments. When I was Georgia's CIO, I was amazed to discover the human services agency had built a data center on its roof (by installing a trailer) just to avoid working with the state's IT department.
The benefits of consolidation in a well designed facility with modern access controls, network perimeter security and operations tools that protect data, as well as facilitate sharing, should be self-evident. But these benefits presume enlightened management. Often a consolidation project uncovers power struggles between the governor, CIO and agency authorities. In these situations, politics overrules technical know-how.
Unfortunately consolidations, no matter how tech-savvy, fail to deliver promised cost savings. Data center consolidation shouldn't be justified as a cost-saving move. In fact, it can only be justified based on improved management opportunities.
Server consolidation is much more complex. Big money can be saved, but the public sector often doesn't understand how these benefits can be achieved. Server consolidation isn't based on cyclical market conditions, but rather on advances in the underlying technology and environmental changes. To understand the benefits, a little history lesson is in order.
First there was the mainframe - essentially a data center wrapped in a metal skin. The big-metal machines benefited both data operators and end-users. As the mainframe market consolidated over time, however, costs rose. Next came UNIX and the open platforms of Sun Microsystems, Hewlett-Packard, DEC and others, opening up a new market, new competition and lower costs.
Microsoft and Intel ushered in the third wave of servers, lowering costs even more. Instead of buying one big server with 32 processors for $500,000, CIOs could get the same computing power by purchasing 16 two-processor servers for $5,000 each. Unfortunately this shift to smaller and cheaper required CIOs to hire more staff to manage the different platforms, which has resulted in a huge, unmanageable mess.
Today the advent of multicore processors, where each piece of silicon does the work of multiple chips, has opened a new architectural era, with blade servers that have the benefit of sharing a common operating system, as well as improved power and heat management.
With this new architecture, server consolidation presents CIOs with a new opportunity. The management of the new environment is tricky, however, and requires truly enlightened operations staffs. CIOs will have to focus on workload management and teaching people to share resources to wring maximum benefits from server consolidation.
CIOs also need to redesign data centers to handle air-cooling and power management. They also need to rearchitect many of their applications to run more efficiently in this new environment. I don't mean migrating the applications from many old small machines into new larger ones. Server consolidation is about real management science: portfolio management and sophisticated resource management.
The key to success is good management and leadership. Too many times, CIOs see consolidation as a technical effort offloaded to data center staff. Wrong. It requires executive effort by CIOs to manage IT resources as an integrated set of assets.
If you're a CIO who takes pride in not being too technical, but rather a "business" manager, it's time to get smart about technology if you want the benefits of consolidation to roost in your data center.
Larry J. Singer is a consultant, and was senior vice president of Sun Microsystems, CIO of Georgia, chairman and CEO of Public Interest Breakthroughs Inc., a research fellow at Harvard University's Kennedy School of Government and an executive director at Texas Instruments.
You may use or reference this story with attribution and a link to