Research On Performance Optimization Of Cloud Service Bus

As Cloud Computing technology develops rapidly in recent years, lots of vendors have launched cloud platforms to provide cloud services. To cut costs, more and more enterprises deploy their application systems in the cloud platform, and then cloud silos emerge consequently. Traditional Enterprise Service Bus is not suitable for cloud integration, thus we put forward the concept of Cloud Service Bus. Based on Enterprise Service Bus, Cloud Service Bus is a cloud platform which provides integration services. As compared to traditional service bus, Cloud Service Bus is confronted with many challenges with respect to performance, security, reliability, etc. Performance issue which arises from large scale applications is especially serious. In this paper, we do some deep research aimed at Cloud Service Bus’s performance.First, we propose a method to infer the network topology in a virtualized environment. Virtualization technology is the supporting technology to construct the cloud platform. However, virtualization has a great impact on network characteristics, so network topology couldn’t be inferred accurately using traditional network analysis technique. We compare three inferring approaches through experiment, and analyze the results. Thereby, we could optimally deploy the integration platform and efficiently integrate the applications in the cloud according to the network topology.Then, we propose to design Cloud Service Bus based on SEDA. SEDA, short for Staged Event-Driven Architecture, is an architecture based on multithreading and event-driven architecture, which could handle high concurrent tasks well. We designed many dynamic resource scheduling mechanism and overload control mechanism, so that Cloud Service Bus could dynamically schedule resources as load varies, and maintain stable performance under extreme overload.At last, we introduce the prototype system of Cloud Service Bus, and design an experimental scenario to show how to integrate cloud service and on-premise applications.