SPGM: an efficient algorithm for mapping MapReduce-like data-intensive applications in data centre network

Xiaoling Li, Huaimin Wang, Bo Ding, Xiaoyong Li

Abstract: In traditional data centre network, how to efficiently allocate the virtual data centres (VDCs) on the physical data centre network (PDCN) is a challenging problem, which is denoted as GraphMap. GraphMap refers to map the virtual nodes to the substrate nodes and the virtual links to the substrate paths, respectively. The existing heuristic approaches attempt a two stage solution by solving the node mapping in a first stage and doing the link mapping in a second stage, which results in the mapping time being very large. In this paper, we propose an efficient mapping algorithm based on shortest path graph matching (SPGM) for online MapReduce-like data-intensive applications; the simulations show that SPGM can efficiently allocate the MapReduce-like data intensive applications on the PDCN in a much shorter time compared to the existing heuristic algorithms and maintain good performance.