Mobility-Aware Serverless Function Adaptations Across the Edge-Cloud Continuum

Abstract

Serverless functions have emerged as a useful abstraction to manage the complexity of distributed and heterogeneous edge-cloud infrastructure. Current cloud-centric orchestration services and serverless platforms are not suitable for the edge-cloud continuum due to their mobility-unawareness. In this paper, we present a mobility-aware framework for edgecloud systems, where the operational mechanisms of placement, scaling, and routing of serverless functions work in tandem. To that end, we formulate the concept of pressure that captures complex system behavior in a single metric. The pressure-based framework handles geo-distributed workload and user mobility by reducing overall function latency and increasing data throughput while efficiently using edge resources. Our contributions include a novel framework revolving around pressure and a realworld proof of concept evaluation. Our approach combines the benefits of a centralized control-plane with a decentralized dataplane. The results show the efficacy of the platform to address operational goals and make effective and deterministic tradeoffs between system utilization and application performance. The novel concept of pressure shows great extensibility and builds the base for a plethora of future works.