Adaptive support to execution of long-running workflows

摘要

Long-running workflow (LRW) is a particular type of workflow whose central aspect is the considerable amount of time to produce results. The long duration of LWRs increases the probability of occurrence of events such as changes in the business rules or interruption of services associated with workflows activities. These events can lead to problems like the loss of intermediary results or the complete reprocessing of activities. The inclusion of adaptive capabilities in LRW execution environments helps to avoid these problems. Existing solutions usually present adaptive skills limited to only replacing services at runtime. In this context, this paper presents E-LONGWisE, an LRW execution environment that incorporates features such as quiescence, data and state management associated with a scenario-based analysis to plan the necessary adaptations of workflows. An experimental evaluation was carried out to assess these adaptive aspects present in E-LONGWisE and demonstrated performance gains running LRWs in the environment.