Research And Implementation Of The Dynamic Flow Control Strategy

Since the rapid development of communication technology,more and more transitional businesses are migrating from offline to online.Meanwhile,the system architecture is switching from monolith style to microservice style.How to effectively manage the growing uncertain user traffic under such complicated situation is an urgent problem which needs to be solved.In this thesis,we researched on dynamic flow control strategy.After studied on transitional flow control algorithms and investigated on existing exploratory researches on dynamic flow control tools.We proposed a hybrid dynamic flow control strategy based on runtime system feedback recognition.First,we analyzed the major challenges of the online service and the classic tools such as limiting algorithms,fuse strategies,and then introduced the current research works on dynamic flow control proposed by the internet industry.Then,based on the deep understanding of the problem domain,we proposed a hybrid dynamic flow control strategy based on runtime system feedback.Our approach considered different use cases.It provided a pluggable mechanism to customize system overload observation method.By default,we suggested to use average wait time of thread pool to indicate system running state.After identify the system status,we pointed out that the further recognition and limitation of the sources of traffic problem would play a critical role in this situation.Finally,we gave the details of the implementation.We tested our approach under several typical problem scenarios.After the experimental verification,our approach was proven to be effective.