Modeling and Verification of Self-Adaptive Service-Oriented Systems

Software has to be continuously adjusted to remain useful (cf. Lehman's law). Otherwise, the software becomes soon useless due to changes of the communication protocols or processes it uses, the physical environment or made assumption about the real world which are not true any more. In addition, software is no longer a separate tool but starts to build up ultra-large ecosystems which elements are connected over networks and therefore will soon reach a size that will be too big to allow stopping or fully replacing it (see ultra-large-scale systems as propagated by the SEI). Therefore, the classical approach to software maintenance and evolution will not work in the future.

We therefore look into means how software can be made capable to adapt its behavior autonomously when required. Such a system is called self-adaptive, as it is able to decide on its own if an adaptation has to take place and how this adaptation is done. In a top-down scenario, there exist one or more top-level components, which decide for all other components what adaptations, are necessary. If a self-adaptive system instead follows a bottom-up approach (in case of simple local rules also often named self-organizing system) each component has the capability to recognize situations, which demand adaptation, and perform the local adjustment. In practice, often mixed forms of bottom-up and top-down self-adaptive behavior are most suitable.

Service-oriented systems are already to some minor degree adaptive systems. Only dependent on the specific service contract definition, the partners look up each other dynamically and take over a specific role. Service-oriented architectures rule the composition of its components via service contracts. The required collaboration (cf. UML Profile and Metamodel for Services (UPMS) RFP) among the different service roles is achieved by two opposed techniques: Orchestration and Choreography. Again we can identify a top-down and a bottom-up approach. Orchestration is the top-down approach and choreography is the bottom-up approach. However, to fully address the outlined challenges for system evolution and maintenance the services have to support self-adaptive system behavior. Without this capability to adjust the system behavior when required also a service-oriented systems would soon become obsolete.

Large-scale software-intensive systems with safety-critical embedded subsystems employing the service-oriented architecture paradigm and self-adaptation are the main application area of our research. These systems are particular challenging. Due to their safety-critical nature guarantees for the proper operation are of paramount importance. However, their large-scale and highly dynamic character does not allow employing existing techniques to provide these guarantees. Therefore, we work on modeling techniques which are able to describe the highly dynamic self-adaptive systems as well as verification technique that are able to cope with these highly dynamic models.