Optimal resource allocation for survivable virtual infrastructures

Abstract

Virtual infrastructures (VIs) consolidated the dynamic provisioning of computing and communication resources. A VI is a set of virtual machines interconnected by virtual links and switches/routers. Infrastructure providers (InPs) manage the physical substrates in which virtual resources requested by VIs (such as CPU, disk, memory, bandwidth) are reserved and allocated. Resource allocation is a complex problem that needs to satisfy different goals: users expect to run their applications on survivable VIs, while InPs aim to maximize profits, minimize costs and reduce substrate fragmentation. However, there is a dichotomy between minimizing substrate fragmentation, by co-locating VIs, and maximizing VI survivability, by sparsely allocating resources in order to decrease the impact of substrate failures. In this context, this paper discusses VI survivability and its impact on substrate fragmentation. We propose a mixed integer linear programming model to allocate resources considering the joint coordination of fragmentation and survivability. Experimental results suggest that it is possible to enhance VI survivability without significantly impacting substrate fragmentation.