Recognizing the Needs of Today’s Resource Hungry Open Systems

Those of us who provided and ran those data centers remember how well technology continually provided for huge dramatic increases in platform performance. The huge increases in technology came at the expense of huge increases in platform footprint requirements, power and HVAC. Bigger meant better and IT people delighted in giving data center tours and showing giant sized water cooled mainframes and vast DASD farms (as they were called) that could occupy as much as an acre of the raised floor.

Then a strange thing began happening. As the years past and hardware technology progressed, those large footprint mainframe systems that we had come to assume would be continually growing in physical size as they got more powerful, actually began to shrink in footprint. It wasn’t all that many years before IT people were proudly showing off much more powerful mainframes with the footprint the size of a refrigerator. Those DASD farms of the past became just another refrigerator size footprint. Along with the dramatic decrease in physical size came an equal decrease in HVAC, power, UPS and generator requirements.

The mainframe platform evolved into a combination of “bullet proof” hardware and a robust and stable operating system. We knew we could look to the future with confidence because with technology we would be able to meet increased user needs within our existing data center footprint, HVAC, power capabilities and technical support staff. As I remember to those of us in IT, our world seemed to be evolving in true homeostasis.

Just when we thought everything was about right with our data center world, open systems, the new kid on the block, began to take over as the user’s choice and began to push the mainframe out. We all knew it was coming sooner or later. Most of us in IT with larger data centers had begun to host some open systems hardware, but never really focused on looking at what the datacenter infrastructure needs were going to be in order to support many of these open systems platforms down the road. Small companies with no data center were simply plugging them into a rack in a closet somewhere. A server’s cost was cheap when compared to the mainframe and so was the software. Chip manufacturers were quickly building new and ever more powerful server chips. More and more new applications were being developed for open systems and became the user’s choice. The IT future with open systems looked only bright and sunny.

Heads up, the datacenter “Tsunami” hits! Along with the increased chip evolution came a wave of some serious growing resource requirements that nobody anticipated. As datacenters like ours added cabinets and filled them with more and more servers, we all began to notice that the available footprint, HVAC, power, UPS and generator resources were quickly being gobbled up. We began to realize that these open systems platforms had insatiable resource appetites when compared to the mainframe. We also began to realize that they required significantly more in the way of unscheduled managed services. This was the opposite of what we thought technology could provide us going forward.

While a server can be quite small in size and amazingly capable, we have found that scalable generally means adding more and more servers in cabinets until we are staring at a large farm of cabinets filled with servers that have a hefty resource appetite and a significant need for those scheduled and unscheduled managed services.

So, what are we learning from experience? And how can it help those solution providers looking to provide them through open systems, as well as those of us running the data centers? Intelligent IT planning and budgeting for open systems support going forward involves a new way of thinking and new rules regarding the datacenter design, HVAC solutions, power utilization, and redundancy requirements and managed services staffing.

The datacenter design for supporting open systems is fast evolving to meet several challenges. We now know that there is a dramatic increase in power to be used as open systems hardware takes over more and more of the data center footprint. This has had a compound affect on open systems actual costs, not only because of the significant increases in what that power costs, but what it means in increased UPS and generator resources required for backing up those increasing power needs.

We also know that several servers in a cabinet do more than utilize power they produce a great deal of heat when compared to the mainframe. Providing cooling directly from underneath a raised floor doesn’t work for servers racked above in a cabinet. Data center designs are quickly evolving to be better able to get cool air from under the raised floor up into the hot spot areas beside each cabinet and also to exhaust the hot air from the other side. Providing adequate HVAC is probably the biggest data center capacity limiting factor today. It is causing many data centers to limit the amount of total power that can be used per cabinet so as to be in balance with their HVAC capabilities.

The open systems environment has led to a whole new service offering for managed services. I remember Gartner warning all of us fifteen years ago, when open systems started gaining traction as a way of the future, that although an open systems platform probably would save on hardware costs, the increased managed services required would probably wipe out the hardware savings. Today many companies choose to look for an outside provider as a solution for providing this service that requires special skill sets.

The end result for all of us in IT from application developers to data center providers is to recognize the need to consider all the realities inherent in today’s open systems environment when planning and budgeting. Those of us who are optimists would like to believe that technology will at some point respond to our power, HVAC, redundancy, and managed services concerns and work toward providing an open systems environment that is not only more robust, but requires less resources and managed services. I think it’s got to happen. It did in the mainframe world. As for future energy costs, don’t ask!

For all of us in today’s world, it’s heads up planning.