In a blog post yesterday, Heather Clancy talked about her conversation with HP's Ed Kettler and discussed a couple of points that he made which are important to consider in any datacenter decision matrix. While I don't necessarily agree with everything Ed said, I do feel that his statements merit further discussion.
Mr. Kettler makes the point that for HP's new datacenter to achieve a high degree of energy efficiency, it is necessary that very specific and tightly controlled equipment and environmental controls need to be in place. Unsurprisingly, this means that in HP's datacenters, this will require the use of HP equipment, since that is what their rack designs call for.
It is only logical that if a datacenter wants to be as energy efficient as possible, and pass the cost savings that represents back to the customer, that the datacenter will dictate to the customer what equipment can be used to meet the datacenter requirements.
When this approach doesn't meet the customer's needs, a flexible (and customer driven) datacenter business model means being able to deliver the appropriate facility model that the customer needs, be it a cage in a large datacenter, to a separate room run under the conditions the customer demands (at an appropriate cost).
Designs for row and rack specific cooling models can be implemented that offer a great deal of flexibility in equipment selection, though I wouldn't expect HP to recommend an IBM solution, or vice versa. If I'm getting datacenter services from a datacenter equipment vendor I would be unsurprised if they express a preference for their own hardware. I'd also be very surprised if the datacenter would turn down a high-value, long-term customer commitment if it required additional equipment from competing hardware vendors to be implemented.
The other point of discussion Ms. Clancy covers is Kettler's "Megawatts to Business value" model, which, briefly stated, is that the cost of keeping specific equipment running may exceed its value to the business. This is a great thing to keep in mind as you design or change existing datacenters. As Ms. Clancy points out, factoring the facilities costs of keeping a legacy application running may change your perspective on the value of that application, and that this value metric is definitely one that should be considered when consolidation projects are being implemented in your datacenters.
As I once again find myself talking about metrics to be used in your datacenter decision processes, I realize that there is a significant metric missing from what is generally discussed. Almost all of the metric discuss power and efficiency; Mr. Kettler's "Megawatts to Business value" starts to move away from the more simplistic power and utilization numbers as decision points and takes into consideration the business operations that energy is being used to support. But what is still missing is a simple "horsepower" rating."
Horsepower is a measurement of a unit of work (though in modern terms it is defined as a specific amount of watts produced). But it is in its traditional sense that I use it here. In your datacenter, how much work is accomplished relative to the amount of energy consumed? And where is that energy being consumed, relative to the work being done?
Utilizing new technologies, such as the latest Intel processors, can allow you to do more work at a reduced energy cost, but that is only a small part of the load. CPU upgrades will show significant value improvement is a "horsepower" metric only if the workload is CPU driven; if your line-of-business applications are storage driven or network bound than the value of more efficient server CPUs is minimized. More careful investments in storage or networking might deliver better business return than simply upgrading servers.
We talk a lot about the energy efficiency of the datacenter, but it is important to keep foremost in your mind that the datacenter serves the business needs and not the other way around. Datacenter design decisions must reflect this across the board.