Low-Latency and Reliable Virtual Network Function Placement in Edge Clouds

Abstract

Mobile edge computing allows efficient cloud servicing by building small-scale cloud infrastructures at the network edge close to end users. The cloud servers implement virtual network functions in a flexible manner without dedicated middleboxes. In such a network design, latency and reliability are two key performance metrics which are not always aligned and a tradeoff can appear between them. In this paper, we identify three key factors that impact the existence of a tradeoff such as the number of services, the number of functions the services require and the existence of redundancy in the implementation of functions. Our main contribution is indicating exactly whether a tradeoff can exist in each combination of these three factors. In case a tradeoff exists, we find guarantees on the performance in a non-optimized metric of an optimal solution to the other metric. We also study the tradeoff when each service should be served in a particular order. We also propose a function placement method that jointly optimizes both latency and reliability. Lastly, we evaluate our analysis in an experimental setting.