Microservice Interface Based Deployment Orchestration

TL;DR

This paper introduces a formal model for deploying microservices, proving the decidability of optimal orchestration synthesis, and demonstrates its effectiveness through simulation on a realistic case study.

Contribution

It develops a tailored formal model for microservice deployment orchestration and proves the problem's decidability, enabling optimal synthesis and run-time global reconfiguration.

Findings

Optimal deployment orchestrations can be synthesized for microservices.

Global reconfiguration outperforms local adaptation in simulations.

The approach is applicable to realistic microservice architectures.

Abstract

Following previous work on the automated deployment orchestration of component based applications, where orchestrations are expressed in terms of behaviours satisfying component interface functional dependences, we develop a formal model specifically tailored for microservice architectures. The first result that we obtain is decidability of the problem of synthesizing optimal deployment orchestrations for microservice architectures, a problem that is, instead, undecidable for generic component-based applications. We, thus, show how optimal deployment orchestrations can be synthesized and how, by using such orchestrations we can devise a procedure for run-time adaptation based on performing global reconfigurations. Finally, we evaluate the applicability of our approach on a realistic microservice architecture taken from the literature. In particular, we use the high-level object-oriented probabilistic and timed process algebra Abstract Behavioural Specification (ABS) to model such a case study and to simulate it. The results of simulation show the advantages of global reconfiguration w.r.t. local adaptation.