Abstract
The problem of time-aware (time series-based) dynamic quality of service (QoS) forecasting has attracted increased attention over the past decade. Developed forecasting approaches have been used to obtain the future values of dynamic QoS attributes for the support of the proactive decisions of various QoS-based applications (e.g., QoS-aware service selection and composition). Thus far, however, a comprehensive investigation and overview of the current research on this topic has yet to be produced. This paper proposes and introduces six assessment criteria which are then applied to the existing literature to produce a comprehensive comparison. Based on this analysis, we describe potential future challenges and research directions in this research area, focusing on gaps in the current literature. This survey provides a clear understanding of the current status of this research area with this paper; additionally, we also technically point out what have to be done by the researchers in this area for the advance of this research topic.
1 Introduction Many software systems rely on external Cloud/Web services to provide required functions or information, and new mobile applications increasingly incorporate existing services (e.g., the RESTful Web APIs exposed by Facebook, Google, and Instagram). The quality of these services (called quality of service, QoS) is an important concern to developers/engineers in service-oriented software engineering (SOSE) and researchers in service computing (SC). Among current QoS-related research topics and concerns, particular attention has focused on the prediction of dynamic QoS properties, such as service response time.
4 Conclusion
Because the values of time-aware dynamic QoS attributes vary over time, considerable research attention has focused on predicting such attributes. The present paper provide a comprehensive survey, comparison, and discussion of relevant studies over the past decade, and proposes challenges and research directions that should be carefully considered for future research efforts.
