Traffic and failure aware VM placement for multi-tenant cloud computing

Abstract

In a multi-tenant cloud, tenants want to receive reliable services and the cloud provider intends to reduce intranetwork traffic in order to provide more services. Achieving the requirements of both sides is a challenging problem. Current tenant abstraction models cannot provide enough information for the cloud provider to optimize network traffic while satisfying reliability requirements. Based on the analysis of the traffic traces of a real multi-tenant cloud, we develop a new tenant abstraction model and design a novel virtual machine placement algorithm, called NETMAP. NETMAP optimizes the cost of network traffic incurred by tenant applications under the reliability constraints from the tenants. Trace-driven simulation results show that NETMAP outperforms a number of possible solutions. Even if there is traffic estimation error from the abstraction model, NETMAP still yields an optimized VM placement.