In-Network Computing With Function as a Service at the Edge

TL;DR

This paper explores the integration of Function-as-a-Service at the network edge, analyzing four models for supporting stateful applications through in-network computing, with initial qualitative and quantitative insights.

Contribution

It provides a comprehensive overview of four stateful FaaS execution models at the edge, including qualitative analysis and simulation-based comparison.

Findings

Four reference execution models for stateful FaaS at the edge.

Qualitative analysis of model advantages and challenges.

Simulation results comparing model performance.

Abstract

Offloading computation from user devices to nodes with processing capabilities at the edge of the network is a major trend in today's network/service architectures. At the same time, serverless computing has gained a huge traction among the cloud computing technologies and has, thus, promoted the adoption of Function-as-a-Service (FaaS). The latter has some characteristics that make it generally suitable to edge applications, except for its cumbersome support of stateful applications. This work is set to provide a broad view on the options available for supporting stateful FaaS, which are distilled into four reference execution models that differ on where the state resides. While further investigation is needed to advance our understanding of the opportunities offered by in-network computing through stateful FaaS, initial insights are provided by means of a qualitative analysis of the four alternatives and their quantitative comparison in a simulator.