Data Center Reliability & Risk Tolerance
- Mar 12, 2009
Sustainability and the environment is something every individual and many businesses have decided should be a priority for the 21st century. With much of the print media’s circulation substantially reduced, due in part to the availability of information via the Internet, through e-books or in other electronic format, we have thankfully started reducing our dependence on paper. But nothing in life comes without a price. With so much of our information now transmitted electronically, we are relying heavily on data centers and their associated infrastructure. According to statistics from the U.S. Department of Energy, data centers are a huge growth industry. They consumed 1.5 percent of all electricity in the United States, according to 2006 data, and power demand was growing at about 12 percent a year, with power and cooling systems operating at less than optimum levels. Who would have thought that Al Gore, who was awarded the Nobel peace prize for his work on the environment, would actually have caused us so much harm when he invented the Internet?! With so much riding on the performance and uptime of your firm’s data center, what is your risk tolerance? Minimizing unplanned downtime reduces risk. In order to design and build a data center that is efficient and does not contribute to the further destruction of our environment, a company must first assess the cost of downtime and its associated risk tolerance. In the past, we were taught that downtime was related to power availability (or unavailability), measured in terms of one nine–90 percent–or six nines–99.9999 percent. For example, for a 24/7 facility, a 90 percent “uptime” or level equates to 87.6 hours of downtime per year. For a facility designed to operate on a 99.9999 percent “uptime,” you would expect 32 seconds of downtime per year. But is any disruption only 32 seconds? What about the costs associated with downtime? These costs obviously vary by industry, but various consultants have estimated raw data and experts have established a range for these costs. In retail, for example, the estimated cost of downtime is about $1.1 million per hour (one would have to assume this is a large shopping chain). On the upper range of the scale is financial services, where the estimate is that one hour of downtime will cost $6.4 million. When I was employed in financial services in the late 1990s, we were always drilled about the possible missed “$1 million trade.” How reliable are these numbers? I would think that if a retail chain had experienced downtime on “Black Friday,” the largest shopping day of the year, or during the holiday season, the $1.1 million figure would probably be two to three times larger. There are actually computer programs that analyze this and establish a firm’s protocol for their information technology design. The more your firm has an aversion to any “downtime,” the more redundancy you will require, driving up the initial construction costs and possibly driving down your DCiE, or data center infrastructure efficiency.In conclusion, the 21st century has brought us many advances with respect to technology and sustainability, but as we were told as young adults, there really is no free ride.Jack Terranova is director of operations and Jack Tucker is director of information technology for Colliers ABR Inc.