Service-Oriented Virtual Machine Placement Optimization for Green Data Center

摘要

The first service-oriented virtual machine (VM) placement for green data center is designed in this work. Integer Linear Programming (ILP) is the problem design basis. The Tree algorithm is proposed to place VM role instances at the lowest communication cost, economizing the construction cost with fewer physical servers. Another Forest algorithm is also proposed for balancing the computation load between the physical machines. Both of the proposed algorithms are formulated on the graph theoretic technique and evaluated and analyzed using the Best Fit algorithm in the simulations. Although the total power consumption and average utility of both proposed algorithms are slightly impaired, the unnecessary outbound communication cost is significantly eliminated and decreased, especially in the immense number of services. The results show that the proposed Tree and Forest algorithms provide lower communication cost than the Best Fit algorithm, and achieve green communications for large scale environment such as cloud or green data center.