Challenges toward QoS-aware service-oriented architectures: real-time performance and dependability

Abstract

One important paradigm in the design of large-scale distributed computer applications is given by the service-oriented architecture (SOA) model. Based on the composition of complex application out of independently developed, loosely coupled, component services, SOA approach relies on the combination of autonomous building blocks that interact coordinataely upon an agreed implementation-independent protocol in order to provide the desired functionality. When deployed as operational components of production systems, novel computer services are supposed to respond synchronously to real-world events associated to the business process they implement, and therefore need to meet temporal constraints dictated by the dynamics of the environment in which they operate. This elicits a real-time system approach. The fulfillment of both the functional requirements concerning the overall business logic and the non-functional requirements concerning the QoS levels expected from the SOA system is a challenging goal in view of the stochastic dynamic of networked computer systems. One emerging concept to cope with such unpredictability of large-scale distributed computer applications is the use of feedback control principles. This research project is aligned along this perspective, focusing on the development of models and techniques aimed at self-adaptive resource allocation in QoS-aware SOA systems with performance and dependability constraints. The investigation addresses run-time control strategies of service selection and request scheduling in composite applications supporting real-time business processes. The study shall result in the definition of methodologies, techniques and tools for enabling self-adaptive QoS provision mechanisms in large-scale distributed systems. (AU)