Optimizing Service Orchestration in OutSystems

Abstract

The growing demand for continuous development and deployment is causing many to steer away from the traditional monolithic architectural style and opt instead for Service- Oriented Architectures (SOAs). Adopting an architecture that is based on loosely-coupled services leads to enhanced modularity and flexibility, further translated into a philosophy of iterative, evolutionary development. The benefits of this development pattern were also made available in the OutSystems low-code platform, with the introduction of services as a new development building-block. Moreover, the independence innate to the multiple services that comprise a SOA hints at possible opportunities for task parallelism: as long as different remote calls to services don’t interfere with each other, they could be performed in parallel. As an immediate result, there could be speedups in multiple parts of an application’s layers. Idle time waiting for data could be reduced, along with internal business logic that could be carried out faster, factors that would positively impact the overall flow of any application. In this thesis we propose the design and implementation of an optimization process that targets the heart of the SOA: the orchestrator itself, the conductor of service interaction that enables the different business processes involved in a software system. The work produced comprises a set of data analysis and representation techniques that work together with the goal of detecting and informing opportunities for safe parallelism in the interaction and composition of the services that make up software factories. The formal definition of the algorithm is accompanied with a prototype that targets the OutSystems platform, with the achievement of considerable speedups in common scenarios. The results obtained suggest the viability of such a mechanism in the world of SOAs.