Joint Content-Resource Allocation in Software Defined Virtual CDNs

The increasing volume, dynamics, and heterogeneity of content distribution services and their requirements challenge the efficiency of traditional content delivery networks (CDNs), in which dedicated storage and delivery appliances need to be pre-provisioned based on estimated peak demands, resulting in excessive over-provisioning and/or degraded quality of experience. A software defined virtual CDN (SDvCDN) is a virtual cache network deployed fully in software over a programmable cloud network infrastructure that can be elastically consumed and optimized using global information about network conditions and service requirements. In this work, we formulate the joint content-resource allocation problem in SDvCDNs as a minimum cost flow problem, whose linear programming solution guarantees service requirements with minimum overall network cost. Our solution jointly optimizes the placement of content objects and the the allocation of storage and transport resources, is applicable to arbitrary network topologies, and captures activation and operational costs, popularity settings, unicast and multicast delivery, as well as capacity and latency constraints.