Control of Server Inlet Temperatures in Datacenters- A Long Overdue Strategy

One of the most critical parameters in the operation of a data center is the inlet temperature of the servers. The ability to control these temperatures is essential for the effective performance of any data center. In this paper we demonstrate why control is important, how it can be achieved, and also present a comparison of the currently available data center cooling technologies, along with an assessment of their control capabilities. We will also show how the Kool-IT technology provides superior control over more conventional approaches, and allows data centers to perform more efficiently, reliably, and effectively. We will further demonstrate that the Kool-IT technology more than pays for itself in both short-term and long-term benefits. Details on efficiency, return on investment (ROI), power density and enhanced server life and performance will also be provided and discussed. INTRODUCTION Data centers have seen an exponential growth in the power density requirements in order to meet the information technology needs of the 21 century. This trend shows no signs of abating and, in fact, has increased faster than some earlier projections (Figure 1). In the 1980’s the power requirements of data centers were low enough that cooling was not a critical problem. The heat sources were not as concentrated as they are today, so only a basic cooling strategy was required. In fact, early data centers did not utilize a hot-aisle/cold-aisle (HACA) cabinet configuration and simply oriented cabinets in the same direction. Over the years, server power densities have kept up with increasing IT demands, however, data center cooling solutions have not. As the power densities grew, so too did the cooling problems, to the point that the overriding problem in today’s data centers is their effective thermal management (Figure 1). According to an April 2005 survey conducted by InterUnity Group, Inc., 59% of data center managers say their greatest stress is the purchase of computing equipment without adequate concern for cooling. To attempt to meet some of the early cooling needs, a few basic innovations were proposed, however, it was not until recently that an effective solution has emerged. Early on, a simple and effective approach was the hot-aisle/cold-aisle design strategy. This was first introduced in 1992 by Robert Sullivan at IBM, at a time when cabinets supported loads of 250 watts/ft or less (based upon the footprint of the cabinet, and not the total square footage of the data center). Today’s cabinets, however, are being asked to support loads of 4,000 watts/ft or more (16 times larger than what it was just 15 years ago) We wouldn’t use 15-year-old server technology in our data centers today, so why are we still using 15-year-old cooling technology? More directly, if HACA designs are so effective, why are today’s data centers experiencing all types of cooling problems? The short and simple answer is a failure of the industry to recognize that we have surpassed the limits of the HACA design strategy, and that we now need to switch to a broader more encompassing cooling solution. With such high loads at the cabinets, the cabinets now need to be considered as an integral part of the cooling system. They can no longer be thought of as passive components, they must now take on an active and pivotal role in cooling a data center. The problem is exacerbated by a misguided belief that all cabinets are the same, and that “(no one) is going to defy the laws of physics. It still takes a certain volume of air at some known temperature to cool a device by a given number of degrees.” [1] Although this relatively 0 2,000 4,000 6,000 8,000 10,00