General background on adaptive systems
Adaptation is an increasingly important characteristic of modern software systems. The objective of autonomic or self-adaptive systems is to provide one or more of the so called self-* properties (self-healing, self-optimization, self-protection) to software systems executing in distributed and highly dynamic environments. Many middleware approaches externalize the adaptation process from the adaptive application itself by applying variations of the feedback control loop model for autonomic systems known as MAPE-K. This model consists of continuously monitoring the system internal state and external environment, analyzing the data to detect undesired operational states, planning how to adapt the system and executing the adaptation plan. All of these tasks are supported by a shared knowledge base. The MAPE-K model separates the concerns of the adaptation process from the application logic, improving the reusability of the middleware solution.
One of the main challenges involved in applying the MAPE-K model for large-scale and decentralized distributed systems is knowledge management: The required knowledge to perform adaptations (context information, adaptation rules, cost functions, etc.) is scattered throughout different and sometimes incompatible models, making it difficult for independent entities (such as adaptation managers and individual control tasks) to share and reuse it. Incompatible communication interfaces, diverging information interpretation and ambiguity are some of the problems indicating the need of a common knowledge base that works as a barrier that abstracts all the apparently conflicting elements of the system. Examples of applications that can benefit form this barrier, are: distributed execution of business processes and scientific workflows, large deployments of wireless sensor networks, distributed interactive multimedia systems, among others.
Master thesis proposal
The focus of this master-thesis topic is on knowledge management for self-adaptive and autonomic systems, and the aim is to study techniques that can be used to semantically represent adaptation knowledge, such as the ontology web language (OWL), and how these techniques can be applied as “abstraction barriers” providing interoperability to heterogeneous and distributed entities. This study is an important step towards defining new ways of designing and implementing the K component of the MAPE-K loop for decentralized and large-scale distributed systems. The main tasks are:
• Study the use of ontology technologies (such as OWL and OWL-S) as a form of specifying adaptation knowledge for heterogeneous and distributed systems;
• Investigate how to apply these technologies as abstraction layers, providing interoperability among adaptation entities;
• Select one of the application scenarios to explore and demonstrate the feasibility and benefits of the proposed knowledge management method.