Slice embedding solutions for distributed service architectures

Abstract

Network virtualization provides a novel approach to running multiple concurrent virtual networks over a common physical network infrastructure. From a research perspective, this enables the networking community to concurrently experiment with new Internet architectures and protocols. From a market perspective, on the other hand, this paradigm is appealing as it enables infrastructure service providers to experiment with new business models that range from leasing virtual slices of their infrastructure to hosting multiple concurrent network services. In this article, we present the slice embedding problem and recent developments in the area. A slice is a set of virtual instances spanning a set of physical resources. The embedding problem consists of three main tasks: (1) resource discovery, which involves monitoring the state of the physical resources, (2) virtual network mapping, which involves matching users' requests with the available resources, and (3) allocation, which involves assigning the resources that match the users' requests. We also outline how these three tasks are tightly coupled, and how there exists a wide spectrum of solutions that either solve a particular task or jointly solve multiple tasks along with the interactions among them. To dissect the space of solutions, we introduce three main classification criteria, namely: (1) the type of constraints imposed by the user, (2) the type of dynamics considered in the embedding process, and (3) the allocation strategy adopted. Finally, we conclude with a few interesting research directions.