In one of the final classes, Professor Gus discussed the problems in scaling up power costs at the same rate that data centers are scaling up to meet the demands of computing power and storage. One measure that was discussed was the PUE, ratio of the total facility power used to the power used for computing and IT equipment. This measurement was brought about by an organization called the Green Grid, “a global consortium of companies, government agencies, educational institutions and individuals dedicated to advancing resource efficiency in information technology and data centers with a holistic approach”. I was interested in the dilemmas facing power usage in data centers, so I did a little research in how efficiency can be increased. The main issues surrounding power usage are the cooling of equipment, which was very evident in the tour of the NCDC, as well as managing the usage of the equipment to make sure it is doing the minimum amount of work it needs to be doing to get jobs done. The efficiency of cooling can depend on many things, including the type used and the sensors that determine how much cooling is needed. There are four types of air conditioning used for data centers, air cooled, glycol cooled, condenser water cooled, and chilled water . Sensors that measure humidity and temperature are also crucial to the efficiency of cooling in data centers. It is important to supply just enough cooling to keep the servers running at optimal temperatures so that no power is wasted over-cooling the servers.
While solutions to efficiency are being worked out and PUEs can be found as small as 1.1 to 1.3, the increases in cloud computing needs are showing that the need for power will increase in the following years to a much larger level. Electricity used by data centers made up an estimated 1.1 to 1.5 percent total energy use in 2010, while in 2005 that number was closer to .5 percent . This number will increase with time, as professor Gus said in class, and alternate solutions to the actual computing may need to be found. One thing I remembered he mentioned was quantum computing, a subject that I was unable to find much information on in terms of applicability to the problems of efficiency. I am interested to see if anyone has any knowledge as to where data centers are headed as far as changes in how computing will be done to increase efficiency rather than how cooling and other uses of electricity in data centers are optimized. Even if the PUE is near 1, data centers still take up a massive amount of electricity, and the changes seem to be needed to be made in the area of how the equipment is using power. Most sources that discuss the issues of data center power usage only focus on increasing PUE, but that doesn’t seem to be where the scaling issues lie.