Real-time energy price-aware anycast RWA in optical data center networks
Arash Deylamsalehi, Pegah Afsharlar, Vinod M. Vokkarane · 2016 International Conference on Computing, Networking and Communications (ICNC) · 2016
A major expense of network and data-center operators is electricity. Geographical distribution of resources implies agreements with multiple utility companies; these are increasingly offering real-time pricing, in addition to standard flat-rate prices. Today data replicas are typically available from several optical data centers. Replicas lead to more flexible network routing. We propose to reduce operational expenditures by choosing the route with the least cost for the energy consumed. We consider price changes throughout the day and the different US time zones and find average improvements of 7% to 11% in total electricity cost and up to 34% in certain scenarios. The greatest benefit of our approach is that the network can dynamically adapt to sudden price spikes or drops; excessive charges are avoided, and price discounts are exploited efficiently. During peak-hours, price differences are typically small and our approach converges to traditional shortest path routing.