Resilient design of a cloud system over an optical backbone

Cloud computing enables users to receive infrastructure/platform/software as a Service (XaaS) via a shared pool of resources in a pay-as-you-go fashion. Data centers, as the hosts of physical servers, play a key role in the delivery of cloud services. Therefore, interconnection of data centers over a backbone network is one of the major challenges affecting the performance of the cloud system, as well as the operational expenditures of service providers. This article proposes resilient design of a cloud backbone through demand profile-based network virtualization where the data centers are located at the core nodes of an IP over elastic optical network. Three approaches, MOPIC, MRPIC, and RS-MOPIC, are proposed. MOPIC aims to design the cloud backbone with minimum outage probability per demand. MRPIC aims to minimize the usage of network resources while routing the cloud demands toward data centers. RS-MOPIC is a hybrid of both approaches aiming to reduce network resource usage while minimizing outage probability. Through simulations of a small-scale cloud scenario, we show that incorporation of manycast provisioning ensures significantly low outage probability on the order of 10-7. Furthermore, integration of a resource saving objective into MOPIC can make a compromise between network resource consumption and outage probability of the workloads submitted to the cloud.