1. Improve use of existing assets
Most data centers underutilize their assets. According to Stamford, Conn.-based Gartner, x86 servers use less than 10% of their available computing power during a 24-hour period. The situation improves slightly for RSC/Unix machines, at 20%. (Brutally efficient mainframes, on the other hand, can achieve utilization rates of 70% to 80%.) Since most of the power and cooling in data centers is sucked up by high-density x86 servers, these underutilized hogs are a big problem. But they are cheap, relative to the cost of software that could allow you to mix and match workloads.
Going through a systematic program of consolidating and optimizing your machines and workloads can push back an energy crisis by months, if not years, Gartner says.
2. Increase the amount of available space
More than 50% of large companies will face space shortages in their data centers during the next five years. That's because the floor space is inadequate to house the projected increase in new servers, and because existing data centers are not designed to power and cool next-generation hardware. So even if the space is there, it can't handle the high-density servers. Building new is complex and costly.
Try extending the use of available space by upgrading the airflow and removing unused cables from the floor. This will enhance air distribution and allow for more servers. But you should look carefully at the cost of renovation, as it can be expensive to change the air conditioning in a legacy data center and fit in chilled systems.
3. Rent, don't own
Another option is to rent floor space for your equipment from a "hosting provider." Most of the sites available have adequate power and cooling and, in Gartner's view, good security. Before you sign the lease, though, you need to figure out whether your organization wants to use the hosting provider long term or as a stopgap measure while it builds anew or refurbishes an existing data center.
Outsourcing your data center is not a strategy to solve your energy problems. The change in delivery of services is too radical a step if the sole motive is to pawn off power and cooling to someone else. And forget about pawn prices. Outsourcers face the same costly energy problems you do, and will charge you accordingly, as they find ways to solve them. Also, as with any type of outsourcing, don't send out a mess. You must benchmark the quality of your services and the cost of delivery, and make the necessary changes to optimize both, before contracting with a third party.
5. Use evolving technologies
The new stuff in power and cooling falls into three categories: hardware designed to manage energy better (e.g., Hewlett-Packard Co.'s BladeSystem c-Class); software to better manage the energy flows within server racks and across data centers; and innovations to data facilities, such as specialty modular cooling (spot cooling, in-row or in-rack). The new technologies show promise, but there are still too few applications out there to make these technologies work together, meaning you will likely have to develop your own software to link them together.Source: "Data Center Power and Cooling Scenario: Options for the Road Ahead," by Rakesh Kumar, Gartner Inc.
This was first published in August 2007