Diana Butnariu, 2013, Computer Science

Diana Butnariu ’13

Companies such as Google, Facebook and Amazon store their data and offer ­information to clients by employing the use of data centers. As the amount of data stored and the amount of information requested, or in other words, the traffic load, to and from data centers increases each year, so does the demand for energy usage inside these data centers. This summer, during my Energy Challenge internship I tried to address and solve the problem of increasing energy usage inside data centers. I first created a small scale replica of a data center in the EDGE Lab and then I implemented an optimal ­server sleeping policy inside the data center replica. This policy aimed to balance energy ­consumption and request handling delays. The technique I employed would save ­energy by putting unused resources or unused servers inside the data ­center to sleep. Initial approaches generated large energy saving, but had the downside of creating equally large delays in client request handling. Thus I took a new approach and implemented a policy that optimally put servers to sleep and woke them up by ­balancing the trade-off between energy consumption and delay in request handling. In order to implement this new policy and test its validity, I replicated a small data center using servers, laptops, and power-meters in the EDGE Lab.