Provisions for Service Co-Evolution - Phase 2 (PROSECCO-2)

点击翻译按钮获取中文摘要

Actively used software must evolve continuously in order to maintain its utility and quality. Adding new features, removing obsolete features, fixing bugs, closing security holes, improving performance – all require updating a software product from time to time. Certainly, this is also true for services provided via a computer network. Support for service evolution is a sine-qua-non for large-scale service environments. The need for service evolution support will grow with the further proliferation and criticality of service-oriented computing systems. Services do not work in isolation. Services have active service providing relationships with clients. Services are part of business processes where services depend on other services. This is particularly true for micro-services. These interdependencies make on-the-fly service evolution a particularly difficult and challenging problem because the evolution of a service may require changes in other dependent services and clients. In analogy to biology, we call this “service co-evolution”. The fundamental research question that we intend to answer in Phase 2 is still the same as in Phase 1: How can we support a coordinated service co-evolution in complex service landscapes composed of a variety of services that depend on each other? In Phase 1, we laid the foundations for multi-party service co-evolution. We defined and classified evolutionary changes that might happen in services and their potential impact on dependent services. We developed an agent-based architecture for service co-evolution that equips every service with an Evolution Agent (EVA) which performs the service evolution in collaboration with other EVAs. Next, we presented a language and communication protocols for communicating changes among EVAs. Additionally, we developed a framework named DiCORE that automates the multi-agent co-evolution process. Last but not least we implemented an initial prototype in the realm of a Smart City scenario. In Phase 2, we will extend the results of Phase 1 and shift our emphasis to the development of techniques, processes and software for the following research objectives: (i) decentralized on-the-fly decision making about the participation and planning of the execution of a service co-evolution for involved EVAs; (ii) monitoring and run-time testing to verify functional and non-functional service properties before, during and after service co-evolution; (iii) on-the-fly optimization of the multi-agent co-evolution configuration and coordination; (iv) an open source prototype implementation and demonstrators in the realm of Smart City. These solutions extend the results from Phase 1 such that in the end we will deliver a comprehensive software infrastructure for service co-evolution.For topic (iii), we will collaborate with Vietnamese partners from two universities in Hanoi.