One of the keys to building an energy efficient datacenter is to minimize energy wasting operations. This might seem obvious, but the ramifications of the process might not be. Beyond looking at power delivery, cooling solutions, and IT hardware that operates as efficiently as possible, there is the issue of managing all of these disparate elements and a single entity. The goal of delivering power, cooling, and computing power only when and where needed in many ways defines the focus for datacenter management in the future.
With this in mind, vendors of datacenter equipment, everything from power delivery units, to chillers, to the actual computing hardware, are instrumenting everything possible. They know that their customers want to be able to monitor and control the workload across their datacenter, and providing that hook, and support for APIs that link into major management platforms will allow potential customers to integrate these products into their existing power management schemes.
In an effort to increase the value-add that their products provide, vendors are beginning to optimize their hardware for on-demand delivery. After all, if you don’t need power, cooling, or a particular IT resource at any given time interval, why waste the energy (and money) to keep it available. So in this vein, the major IT vendors are offering up APIs for their management tools designed to take advantage of this level of control of the datacenter infrastructure. And therein lies the rub; to properly implement and maximize energy savings with a completely on-demand infrastructure every controlled component needs to behave in an expected, and predictable, fashion.
This means that the management infrastructure needs to be tightly tied into the physical infrastructure. The haphazard addition of software and hardware, or even well-planned additions, can wreak havoc on the on-demand nature of this prospective datacenter. This is especially true as vendors add their own flavor of on-demand capability to their hardware. For this concept to provide maximum value, behavior across the hardware needs to be consistent. And as products are added to the infrastructure, they have to work with the on-demand nature of the infrastructure without causing problems.
This is a step beyond advertising that your product supports a given vendors management or on-demand API functions; it means that the product has to be sufficiently well integrated with the buyers management infrastructure to be able to deliver on its promise. And it’s that integration that will make or break the product.
This is where the single-vendor solution will really shine, be it the “we sell just about everything” approach that vendors like HP are offering or the pre-tested multi-vendor partnership configurations of which Cisco is the leading proponent. These vendors are going to be able to offer tested configurations with documented energy use statistics, for a complete datacenter infrastructure. As factors such as PUE become the standard metric for evaluating datacenter efficiency (and hence, the purchase of the equipment therein) the ability to deliver a proven efficient datacenter infrastructure will be a very compelling selling point.
