Service-Oriented Computing (SOC) is based on services as the primary constructs for rapidly developing flexible and dynamic software at a relatively low cost. Services are platform-independent computational entities which can be combined to fulfill application requirements in a platform independent way. However, current service standards and technologies do not yet provide sufficient means for developing adaptive service-based architectures capable of adequately handling the dynamism of contemporary service-oriented software ecosystems.

In order to adapt to this dynamism, atomic services are commonly combined to form composite services of higher granularity and richer functionality. Resulting composite services can be used as constituent services in further service compositions or they can be offered as final solutions to service clients.

We refer to dynamic services (self-adaptation, self-configuration, self- healing, self-optimization) by the term polymorphic, since the functionality they provide can vary depending on their status and environmental context.

Given the above context, the aim of the proposed research is to contribute towards a framework for autonomic adaptation and composition of polymorphic services. The main objectives are:

  • Investigation of modelling principles and notations for representing polymorphic service adaptation and composition, identifying “gap” areas where existing work does not sufficiently support the desirable autonomic features of polymorphic services.
  • Development of an agent-based reference architecture implementing mechanisms for polymorphic service composition and management based on bio-inspired self-organisation and decentralised control models, and business value criteria. The architecture would synthesise existing results and develop original ideas in the “gap” areas identified above.
  • Integrate the architecture within a cloud-based large-scale experimentation infrastructure, allowing the tuning of key parameters of the proposed framework through experimentation. The experimentation will take place whilst creating the framework, for establishing the range of parameters for different application domains. However, a link to a real-time connection to the cloud during run-time will also be provided, allowing the offloading of computationally intensive calculations and on-line simulation mechanisms to the cloud.
  • Evaluation of the proposed approach (how well the autonomically composed and managed polymorphic services satisfy the requirements of their specific domain). The evaluation will be divided into quantitative, using results from simulation experiments modelling realistic scenarios, and qualitative, using opinions of key stakeholders gathered through semi-structured interviews.