Probabilistic reliability prediction models for task scheduling in distributed systems: A review

In service-oriented distributed systems, beside time and cost, reliability is the most important concern to both service users and the service providers. Although, this has been many decades problem, the existence of large number of service systems on the internet today has rendered the problem more difficult. This is because the distributed environment of today is more complex with numerous uncertainties and chances of failure at all levels. Therefore, selection of reliable service poses a serious challenge. To combat this problem, over the years, huge number of reliability researches has been reported in literature. These researches have been categorized and analysed in many survey and review studies. However, most of these studies focus on the architecture-based reliability mechanisms and pay little attention to the advances in the popular probabilistic reliability prediction methods which are based on quantitative reliability measurements. These methods which are sometimes called ‘black box’ techniques are of great importance to both service designers and service clients such as brokers and other proprietary schedulers, for evaluating reliability of services or service components. Therefore, in this study the previous survey and review studies are extended by analyzing these methods and their recently proposed variants. In the end the study reveal some of the current issues that need further research