Prediction of Non-Functional Properties of Service-Based Systems

Abstract

From the outset, service-based systems offer several advantages, including the promise of shortening the development life cycle, reducing the costs of software development and faster utilisation of recent technical improvements in the software industry in terms of capability, reliability, compatibility, performance, and so forth. However, this is rarely realised in practice. An important issue of service-based systems is that they are likely to be (or perhaps already being) used in domains where human life and/or economic loss are possible and the need for a highly reliable system is a must. However, would it be possible to build reliable service-based systems that meet such domains’ requirements? Does it necessarily mean that composing a system from highly reliable services produce a highly reliable system? Is it possible to predict quality attributes of a service-based system from its services before building it? In this chapter, the complexity of this issue is highlighted, focusing on reliability, in an attempt to answer some of these questions. The chapter outlines various approaches that attempt to address this issue, and proposes a possible way forward for predicting the reliability of service-based systems from its individual services.