Using Decomposition and Local Search to Solve Large-Scale Virtual Machine Placement Problems with Disk Anti-Colocation Constraints

摘要

This paper addresses the problem of consolidated virtual machine (VM) placement in cloud computing taking into account the disk anti-colocation constraint. For every VM assigned to a physical machine (PM), this constraint imposes that the VM virtual disks should be spread out across the PM physical disks. For scalable solutions, the proposed algorithm combines decomposition method with local search heuristic. The proposed approach outperforms existing approaches in operation cost and in execution time (or convergence time) for large scale instances.