Optimization of Network Services Embedding Costs over Public and Private Clouds

Abstract

Public cloud providers offer individuals and businesses the ability to rent IT resources to meet their needs without investing in their own hardware. At the same time, the Network Functions Virtualisation (NFV) concept promotes the migration of network operators from expensive and poorly scalable hardware network devices to virtual softwarized network functions. In order to embed those functions, network operators may decide to subscribe to public cloud offers. However, their diversity, both in terms of resource capacity and price, makes it difficult to find the optimal combination of offers that meets all needs at the lowest cost. In this paper we propose to solve this issue with an algorithm designed to help a network operator to select the best combination of offers (in terms of price) to reserve the virtual machines needed to support a set of network services. We analyze the computation time of our solution against various metrics, and estimate the cost savings compared to a traditional resource provision scheme or an unplanned resource rental strategy. Finally we evaluate the opportunity for a network operator to build its own datacenter, considering the existence of offers from public clouds.