A sampling-based online Co-Location-Resistant Virtual Machine placement strategy

Abstract

In this paper, we discuss the co-location attack problem in the cloud IaaS from the Virtual Machine (VM) placement strategy perspective. We formulate the online secure optimization VM placement problem in a way that guarantees—apriori—a specified level of security while minimizing the number of used physical servers. To solve such a problem, we propose an approximate online secure VM placement algorithm based on sampling. The polynomial-time algorithm is based on a sound security inference procedure based on the confidence interval estimation method. Our empirical results demonstrate the correctness and the effectiveness of our approach in guaranteeing a Co-Location-Resistant (CLR) VM placement with a specific level of confidence and a threshold error as new incoming VM requests are being assigned to servers online. We compared our algorithm to a CLR alternative presented in Azar et al. (2014).