Key Selection Strategies for a Novel Service Index Model to Expedite Service Discovery and Composition

The rapid growth of Internet applications and the widespread use of cloud computing platforms have accelerated the development of Service Computing. As a result, the number of services on cloud computing platforms is growing exponentially, and these services are often provided by different providers for building more complex applications to meet the needs of users. However, due to the diversity of services, a single service is often unable to meet the needs of users, and they need to be used in combination. This poses a major challenge for the management and discovery of Service Computing. To address this challenge, the multilevel index model has become the state-of-the-art approach for managing and retrieving services from service repositories in the Service Computing domain. However, adding and retrieving services to the model in a timely and accurate manner remains a persistent problem. These problems affect the efficiency of discovering and composing services in Service Computing, and thus, new methods and techniques need to be investigated to optimise the management and discovery of Service Computing.

The existing key selection methods for solving this problem do not effectively solve the problem of service addition and are based on the assumption that the distribution probability of service parameters is equal, which is not always the practice case. Certain services have the same input or output parameters, and some are frequently invoked by users, resulting in an unequal invocation of retrieval request parameters for each service. Furthermore, the invocation frequency of popular services changes over time. Existing key selection methods cannot handle such changes, thereby cannot guarantee service retrieval efficiency. In light of these challenges, this thesis aims to optimise and enhance the multilevel index model. Suitable key selection methods have been proposed to address issues including the inefficient service addition operation, equal appearing probability of service parameters, and service hotspot drift respectively. By reducing the time incurred during the service addition and retrieval process in the index model, the proposed approaches can efficiently improve the efficiency of service discovery and composition.