GreenMap: Green mapping of MapReduce-based virtual networks onto a data center network and managing incast queueing delay
Energy consumption is a first-order concern for today's data centers. MapReduce is a cloud computing approach that is widely deployed in many data centers. Toward virtualizing data centers, we consider MapReduce-based virtual networks that need to be embedded onto a data center network. In this paper, we propose GreenMap, a novel energy-efficient embedding method that maps heterogeneous MapReduce-based virtual networks onto a heterogeneous data center network. Besides, we introduce a new incast problem that specially may happen in Virtualized Data Centers (VDCs). GreenMap also controls the incast queueing delay. We formulate a Mixed-Integer Disciplined Convex Program (MIDCP) for this method. Because the formulated MIDCP is NP-hard, we also propose a novel and scalable heuristic for GreenMap. Simulation results prove that both of the MIDCP and the heuristic reduce a data center network's energy consumption effectively, and control the incast queueing delay.
|Keywords||Energy-efficient data centers, Incast queueing delay, MapReduce, Virtualized data centers|
Ghazisaeedi, E. (Ebrahim), & Huang, C. (2017). GreenMap: Green mapping of MapReduce-based virtual networks onto a data center network and managing incast queueing delay. Computer Networks, 112, 345–359. doi:10.1016/j.comnet.2016.11.015