Energy-Efficient, Green Job scheduling for Geographically Distributed Data Centers

Abstract

Geographically distributed data centers are used as backbone infrastructure by big IT companies to meet rapidly increasing IT services demands and as the use of internet and information technology is increasing across the globe. High energy consumption,increased operational expenditures and high car- bon footprint are becoming points of great concern related to data centers. In this work we present green aware, network aware, energy e cient job scheduling mechanism for geographically distributed data centers.It consid- ers availability of green energy at each data center to maximize utilization of green energy, it considers the amount of under utilized computation resource at individual data centers while assigning job to a data center, which helps to achieve better server consolidation that results in better energy e ciency and also considers network load at each data center which helps avoiding hotspots in data center networks. This approach to data center selection mechanism helps to create better balance between server consolidation and network utilization at individual data centers. In this work we also present a framework for simulating infrastructure of geographically distributed data centers based on OMNET++, an open source object-oriented modular dis- crete event network simulation framework. It also contains power consump- tion model for calculation of energy consumption and a hierarchical statistics collection model for performance evaluation. We have implemented and eval- uated the performance of our job scheduling mechanism on this framework to prove e ciency of our job scheduling mechanism.