Building high-speed roads: Improving performance of SOAP processing for cloud services

Abstract

With the advent of cloud computing, more services should be delivered from a cloud or datacenter to clients via Internet. Web Services has many mature specifications and has become a popular mode to support cloud application. However, the performance has been a key character to guarantee the quality of remote service in a cloud. In particular, the Simple Object Access Protocol (SOAP), being based on Extensible Markup Language (XML), inherits not only the advantages of XML, but its relatively poor performance. This makes SOAP a poor choice for many high-performance Web Services. Experiments have shown that the XML parsing, formatting and type mapping are the primary performance bottlenecks of SOAP processing. In this paper, we proposed a new high-performance SOAP processing architecture named HiSOAP which is based on several techniques such as on-demand building, delayed-databinding and adaptable services invoking mechanism. HiSOAP has the benefits of reducing XML object model, saving memory and eliminating Java reflection. We have implemented the SOAP engine HiSOAPprototype, and the experimental results show that the delayed-databinding and eliminated-reflection techniques improve the performance of SOAP processing dramatically, and it out-performances Axis2, especially in the complex compound data types of Web Service.