Moore’s Law states that at our rate of technological development in the semiconductor industry, the number of transistors doubles every year (now every 18 months). As the number of transistors increases, processing speeds go up, which results in greater heat loads in the same area. The result: Your computer rooms are getting smaller.
The new rack and blade servers are much more efficient in terms of processing speeds and capacities. They take up less room, are easier to cable, and components can be changed out much easier than with older legacy systems. However, the heat load is going up in terms of square footage. Some racks can have up to 24 kW in a 7-square-foot area. This translates into almost 7 tons of cooling, or, put another way, 1 ton per square foot of equipment.
Many small- to mid-sized IT centers are designed using traditional rules of thumb. In a worst-case scenario, the design consists of: “We’ve only got $50,000 in our budget for HVAC. What can we get with that?” The reality is that a system designed incorrectly is a waste of capital funds and energy. A system that is undersized will provide inadequate temperature and humidity control, and either issue can affect reliability.
The steps in developing a proper HVAC system for today’s IT centers usually take one of two forms: First, the owner hires a consultant, plans are developed, and bids are taken. The low bidder is awarded the contract. The second alternative is having a design/build contractor develop a proposal. The ins and outs of the project are negotiated and a price is agreed upon. Design/build projects are often fast-tracked and will result in a price at or below the plan and spec price.
Professor Dean Kashiwagi of the Del E. Webb School of Construction at Arizona State University, Tempe, AZ, has astutely pointed out that a specification is viewed as a floor by the owner and/or consultant. The contractor views the specification as a ceiling. The difference between the two is how change orders and lawsuits can result.
The most important thing to look for when developing a proper HVAC system is a person who has experience in IT centers. Next, review what the heat load will be from the IT supplier. Back when the term “360 computer” meant an IBM mainframe, there were site preparation guides. These guides listed virtually every component necessary to prepare the data center for operation. The result was that large data centers had fewer problems during operation.
Today’s IT centers are designed with various suppliers’ equipment. HP ProLiant™ equipment sits next to a Dell PowerEdge™ blade server. Some system integrators do not have the technical expertise to develop a cooling load; their expertise is computers, not HVAC. Equipment manufacturers have the BTU output of every piece of equipment. The system integrator or IT manager must make the gathering of this information as important as finding out the number of gigabytes used for storage. If the system gets too warm, there is deterioration in processing speed and reliability. As an example, The Uptime Institute, Santa Fe, NM, has reported that equipment in the top-third of a rack fails twice as often as the equipment in the bottom two-thirds.
Proper pre-planning for IT centers can result in affordable capital budgets, optimum energy utilization, and better equipment reliability.
Rich Goulet is president of Goulet & Associates, an HVAC operations and maintenance consulting firm. He can be contacted at ([email protected]).