Workloads in data centers are highly dynamic. Secondly, data centers are often hybrid in terms of hardware/compute capability. An option at one extreme is to keep all machines running all the time – the downside is that cost of running (including energy cost) becomes very high. On the other hand, if workload is broken down into review points (say every 5 minutes), then based on the aload some servers may be put in sleep mode (to reduce energy cost); but this then bring up an additional cost facture due to machine turn-on/off, which for example reduces the hardware lifetime. Thus, it is important to understand how to optimally use resources in a data center so that the cost can be minimized over a time window. It may be noted that it may be possible to merge review points to extend the span of review points, if the workload doesn’t change drastically, or we decide to aggregate them to potentially reduce cost. However, this is not as simple. Due to aggregation, the energy cost may actually go up.

 

Thus, we have considered the problem of balancing energy consumption and  system cost in hybrid data centers with dynamic temporal workloads. For comparison, we consider three data-center scenarios: all homogeneous (i.e., all machines are of the same type), all heterogeneous (each one  different), mixed clusters different clusters, where each cluster has homogenous nodes). Specifically, for each data center type, we’ll present time dependent optimization models that capture the workload requirements as well as the different cost factors.

 

Through our study, we found that the computational time for the heterogeneous model is most time consuming. Thus, to cut this down, some aggregation is necessary; on the other hand, as point out earlier, this increases the energy cost. Thus, we will discuss the trade-off between energy-aware aggression and the computational cost, and where and how a dynamic aggregation scheme helps compared to a static aggregation scheme. 

 

Bio: Deep Medhi is Curators' Professor in the Department of Computer Science and Electrical Engineering at the University of Missouri- Kansas City, USA. He received B.Sc. in Mathematics from Cotton College, Gauhati University, India, M.Sc. in Mathematics from the University of Delhi, India, and his Ph.D. in Computer Sciences from the University of Wisconsin-Madison, USA. Prior to joining UMKC in 1989, he was a member of the technical staff at AT&T Bell Laboratories. He was an invited visiting professor at the Technical University of Denmark, a visiting research fellow at Lund Institute of Technology, Sweden, and a Fulbright Senior Specialist. He is the Editor-in-Chief of Springer’s Journal of Network and Systems Management, and is on the editorial board of IEEE/ACM Transactions on Networking, IEEE Transactions on Network and Service Management, and IEEE Communications Surveys & Tutorials. He has published over a hundred papers, and is co-author of the books, Routing, Flow, and Capacity Design in Communication and Computer Networks (2004) and Network Routing: Algorithms, Protocols, and Architectures (2007), both published by Morgan Kaufmann Publishers.