Operating large-scale IoT systems through declarative configuration APIs

Abstract

Configuration activities constitute a large part of the work in operating an IoT system. Such activities include the onboarding of devices and rollout of firmware updates. Configuration activities must be carefully vetted to prevent faulty states, putting a lot of pressure on IoT system operators to get it right. The problem is further exacerbated due to the heterogeneity, scalability, and distributed computation challenges typical of large-scale IoT systems. Therefore ensuring safe and reliable operations requires careful design. In this paper, we argue that a declarative approach to configuration, complemented with a discovery-driven API design is ideally suited to solve these challenges. We present suitable abstractions needed to realize such a declarative configuration API. Our experiences show that the proposed abstractions and API model are well suited for the purpose of large-scale IoT systems, and allow for high degree of safety and reliability.