Decentralized learning for self-adaptive QoS-aware service assembly

Abstract The highly dynamic nature of future computing systems, where applications dynamically emerge as opportunistic aggregation of autonomous and independent resources available at any given time, requires a radical shift in the adopted computing paradigms. Indeed, they should fully reflect the decentralized perspective of the execution environment and consider QoS, scalability and resilience as key objectives. In this context, the everything-as-a-service (XaaS) paradigm, which envisions the creation of new services as an assembly of independent services available within the environment, can greatly help in tackling the challenges of developing future applications. However, in order to be effective, XaaS paradigm requires self-adaptive service assembly solutions able to cope with the unpredictable variability and scalability of the execution environment, the lack of global knowledge, and the QoS requirements of services to be built. We contribute in this direction by designing a fully decentralized and collective self-adaptive service assembly framework whose main features are: (i) self-assembly, i.e., the ability to operate autonomously, (ii) online-learning, i.e., the ability to dynamically learn from experience, (iii) QoS-awareness, i.e., the inclusion of QoS requirements as driving forces for self-assembly, (iv) scalability, i.e., the ability to cope with a large number of services, and (v) resilience, i.e., the ability to maintain the persistence of service delivery when facing unexpected changes (e.g., in the number and/or QoS of services). Simulation experiments show that our solution makes the system able to quickly converge to viable assemblies that improve and maintain over time the social welfare of the system, despite the local perspective of each participating service.