Intent-based Management for the Distributed Computing Continuum

Abstract

Managing digital and connected systems has become increasingly challenging in the past decade due to their scale and complexity. A new perspective is required to manage these systems, considering the infrastructure and components from edge to cloud, i.e., in the distributed computing continuum. Serverless computing offers improved scalability and cost efficiency, but balancing and coordinating serverless systems remain complex. Intent-based systems, popular in networking, can provide a solution by translating stakeholder inputs into actions that meet Service Level Objectives (SLOs). Their application in the computing continuum can be highly beneficial, but it has yet to be deeply explored. To bridge this gap, we propose a methodology for deploying an intent-based system for the computing continuum. We implement an architectural framework leveraging the serverless paradigm. Furthermore, we focus on defining and implementing the main components for translating the management requirements into actions executed by serverless functions inspired by a three-layer model. Through a Proof of Concept (PoC) deployed in Amazon's AWS cloud and detailed simulations, we showcase how such an approach can resolve conflicts in a complex system, i.e., balancing efficiency and availability. Our work aims to contribute to effectively managing the computing continuum and highlight the potential of intent-based systems in this domain. The experiments' results show our framework's ability to make appropriate scaling decisions, fulfilling both objectives.

Topics & Concepts