Data-Centers and Cloud Computing


Clouds are growing in importance as more and more services are being migrated into them. With the adoption of the cloud as a model of deploying services, comes new challenges in managing these services. In this work, we examine the underlying data centers within these clouds. In studying the data centers, we aim to understand the characteristics of traffic within the data center and to use our observations to develop a traffic engineering technique for alleviating loss and reducing congestion. From our study of data center traffic, we observe that traffic at the edge is ON-OFF and this ON-OFF arrival process can be parametized with a set of heavy tailed distributions. In addition to this, we observe that a significant portion of the traffic matrix is predictable at short time scales. To leverage these observations, we propose MicroTE, a traffic engineering techniques that reduces congestion by routing predictable traffic around hotspots within the network.



Data, Source code


Graduate students: Ashok Anand, Theophilus Benson.
Collaborators: Sambit Sahu, Anees Shaikh, Ming Zhang.