Where to fix temporal violations: a novel handling point selection strategy for business cloud workflows

On-time completion is an important QoS (Quality of Service) dimension for business cloud workflow systems. Due to the dynamic and uncertain system performance, failures of on-time completion, namely temporal violations, often take place during workflow runtime and thus temporal verification and temporal violation handling are required. Currently, workflow throughput is employed by temporal verification for a large number of parallel workflow instances as the detector for temporal violations in a collective way. However, based on workflow throughput, it is impossible to locate the exact workflow activities and cloud services where temporal violations occurred. This is a critical problem as to increase workflow throughput of the whole system is very expensive if we do not know where to fix local temporal violations. To address such a problem, based on the results of a throughput based temporal checkpoint selection strategy, this paper proposes a novel temporal violation handling point selection strategy which further utilizes the average response time of a cloud service as the key detector to determine whether temporal violation handling needs to be implemented or not. Comprehensive experimental results demonstrate that while achieving the same on-time completion rate, our strategy is much more cost effective as it can precisely localize temporal violations and hence require a much smaller number of times for temporal violation handling compared with other strategies.