by Paul Rudo on 05/04/11 at 2:23 pm
Dimension Data is a specialist IT services and solutions provider that helps clients plan, build, support and manage their IT infrastructures. They were founded in 1983 and headquartered in Johannesburg, South Africa. Currently, Dimension Data operates in 49 countries across six continents.
Today, I’ll be interviewing Kris Domich, who is the principal data center consultant for Dimension Data Americas.
What sorts of problems are companies facing when it comes to the power requirements of their data centers?
Most data centers have or will experience constraints on how much conditioned power they can provide to some portion or all of the racks. This is commonly due to the increasing power densities of late model equipment and the adoption of such technologies without properly planning and anticipating the need for an adequate power distribution system.
What are some of the biggest mistakes that companies which leads to inefficient data center power usage?
The most visible mistakes tend to be overcooling, or overuse of the wrong type of cooling. An inefficient cooling strategy will lead to the deployment of more cooling components than may actually be required — this will drive use of unnecessary power. Other examples include poorly planned space configurations or ones that were initially planned well but degraded over time. Lastly, a lack of a capacity planning and management regimen will also lead to energy and space inefficiencies. Much of this can be remedied by gaining alignment between IT and facility departments, and ensuring that capacity planning and management is a joint effort between both groups.
What are some tips that you can give when it comes to designing an efficient data center?
Understanding what critical loads must be supported over the planning horizon of the data center is paramount when designing power distribution and cooling subsystems. It is also critical to understand the availability requirements of your organization. This will allow you to choose the right amount of redundancy and reduce the risk of over-provisioning and driving inefficiency. Organize the physical data center based on load profiles, and if possible, create an area that is designated for higher density equipment and other areas that are not. This will allow you to prescribe the proper cooling regimen for the various equipment types and lessen the chance that the entire data center is designed to support a single power density – resulting in over-cooling and potentially less cooling that some of your equipment may require.
What are some good best practices that companies can adhere to when it comes to implementing heating and cooling systems within their data centers?
Baseline critical load today and work with IT to develop a growth plan over at least a five year horizon. Develop notional power and cooling designs that meet the requirements over the entire planning horizon. Translate those designs to specific technologies that are modular enough to be expanded beyond the planning horizon.
What are some of the most efficient data center heating and cooling systems? On what criteria do you base this decision?
Precision and variable output systems are the most efficient. These systems are designed to provide cold air where it is needed (i.e. at the equipment air inlets) and evacuate hot air from its origin (i.e. equipment exhaust points). This precision design is more efficient than the traditional means of “flooding” the room with cold air and having no deliberate means to evacuate heat. The variable aspect of these systems means that as loads increase and recede, so does the amount of cooling supplied, which results in a more efficient use of energy.
How can companies get a better idea of their data center resource utilization?
Implementing a meaningful set of instruments and paying attention to what they tell you will provide intelligence on a number of levels. Such instrumentation can initially provide visibility into what loads are being generated by specific equipment at a given time. This intelligence can aid in meeting the needs of nominal loads and not designing to support maximum loads 100% of the time, which is a rare occurrence.
What are some common power-hungry services which could easily be made more efficient?
On the average, the cooling subsystem — including everything from the chiller/condenser through to the computer room air conditioners (CRACs) — is among the most power hungry subsystems in the data center. They can be made more efficient through the use of precision and variable-output components. This will ensure that the cooling subsystem is not operating at levels which exceed the amount of critical load being generated by data center systems.
What kind of cost savings can a company expect from refining the efficiency of their data centers? Are there any other benefits besides cost savings?
Actual cost savings and avoidance will vary by organization and vary by the degree of inefficiency that is corrected. Historically, we have seen reductions of overall power consumption in excess of 30% and as much as 60% or more. Another way to look at savings is to examine the cost to run a given workload. When introducing efficient power and cooling strategies, organizations can more easily capitalize on more efficient computing platforms such as blade servers.
Adding virtualization to these platforms can dramatically reduce the cost to run a given workload and allow for the running of more simultaneous workloads. As a whole, the IT “machine” becomes more efficient and can respond to business requirements faster.