High Availability Solution for Virtualized Local Disaster Recovery

Abstract

Disaster recovery (DR) is an important element of the complex information technology (IT) systems. The availability of the IT for everything, from everywhere, at all time is a growing requirement. Effective IT strategies need to have both high availability (HA) and disaster recovery (DR). Nowadays, virtualized platforms have become the most popular option to deploy complex enough services. Software availability is one of the weakest links in system availability. Web servers have continuous execution of long duration and with rather varied workloads. Such characteristics make them potential candidates for a degenerative phenomenon called software aging. The work presented in this paper aims to offer the high availability solution against software aging of virtualized local disaster recovery (VLDR) by providing measurement based software rejuvenation. The idea behind our paper is two-fold. First, we present the framework seeks to maximize the number of services running simultaneously, while guaranteeing the resources needed by each service. Second, we estimate the time to aging-related failures and then which used as aging failure rates for measurement based software rejuvenation through a stochastic reward nets model. Finally, we perform the numerical analysis to evaluate the performance of the model.