Towards dynamic reconfiguration of composite services via failure estimation of general and domain quality of services

Abstract

Composite services (CSs) are large-grained services composed of simple ones to satisfy users’ complex requirements. A major issue concerning a CS is that the general quality of service (GQoS) and the domain quality of service (DQoS) of its component services often fail due to environmental changes. However, existing research fails to consider how to actively perceive GQoS failure and DQoS failure of component services, and cannot perform dynamic reconfiguration of composite services proactively. To tackle this problem, we propose an approach for dynamic reconfiguration of CSs by automatically detecting both GQoS and DQoS failures. Our approach includes three stages: at the first stage, we estimate the failure judgement basis of GQoS and DQoS for each component service based on a preferences-aware global GQoS decomposition method and a statistical method. In the second stage, during the execution of a CS, we proactively estimate whether the component services to be executed will encounter GQoS or DQoS failure based on a novel GQoS and DQoS joint prediction method. At the third stage, when the component service to be executed encounters GQoS or DQoS failure, we reconfigure the CS by replacing the failed component service with a best-matched similar service, while we also apply the GQoS and DQoS joint prediction method to predict GQoS and DQoS of the matched service, so as to ensure that the matched service can satisfy the local GQoS and DQoS constraints of the failed component services. Moreover, to improve the efficiency of replacement services selection, we cluster the large scale services based on a multi-attributes aware services clustering method in advance. Extensive experiments have been carried out to demonstrate the effectiveness and performance superiority of our approach.