一种主动式的网格工作流可靠性保障方法

A Proactive Approach to Ensure Dependability of Grid Workflows

针对中国国家网格(CNGrid)环境,曾尝试通过一种工作流元调度机制(VINCA抽象工作流),在为用户提供单一入口和屏蔽细节的同时,优化利用已有的流程引擎能力。在此基础上提出一种主动式的工作流可靠性保障方法,根据流程引擎在最近一段时间间隔内的失效率和负载增长率两种特征参数主动预测其将来成功处理请求的概率,并据此将VINCA抽象工作流中的复合活动(实现为一个子流程)调度到"最可靠"的工作流引擎上。文章旨在从整体上提高工作流执行的成功率和稳定性,有效地避免基于"事后"被动恢复模式所带来的时间开销和实施上的复杂性。最后,通过场景示例作出了定性分析,表明该方法在大规模持续执行流程时,能充分利用工作流引擎能力,有效地保证工作流执行的可靠性。

As a special kind of ＂programming＂ technology for constructing problem-solving applications on the basis of grid resources, grid workflow has attracted attention and made progress. However, how to ensure dependability of grid workflows is still a remaining challenge. For China National Grid （ CNGrid）, a meta-scheduling mechanism （VINCA Abstract Workflow） was proposed, by which the underling capability of workflow engines can be optimally used and a single entrance can be provided while details are hidden. A proactive approach to ensure dependability of grid workflows is proposed in this paper. The workflow engine＇s failure rate and workload increasing rate in a certain recent interval are calculated for predicting the probability of successfully handling current execution request. The compound activity （implemented as a sub-process） in a VINCA abstract workflow is scheduled to the most promising engine. In this way, the overall dependability of workflow execution can be enhanced while avoiding time cost and technical complexity caused by the traditional ＂reactive＂ rescue approaches. The approach is qualitatively analyzed by an example scenario, which indicates that the approach can ensure the dependability of workflow by fully utilizing the engines＇ capability when executing workflows cosmically and continuously.