Wire-Speed Differential SOAP Encoding

Abstract

The use of SOAP based communication protocol in complex Service-Oriented Architecture (SOA) environments can cause a significant strain on network resources thereby degrading application level performance. In this paper, we present software and hardware approaches for transparent linerate encoding and decoding of SOAP traffic for alleviating network bandwidth bottlenecks. The proposed approaches are significantly lighter weight than gzip and binary XML encoding techniques while achieving better compression ratios. First we present a software-based scheme with a linear runtime that uses a finite set of pre-selected message templates for differential encoding of SOAP messages. We propose generating the templates by extracting the longest common subsequence from a representative set of SOAP messages encountered during the service operation. Next we present an architecture for line-rate implementation of data differencing functionality in hardware. As compared to the software-based scheme, the hardware design allows higher flexibility in the template selection and message encoding without compromising the encoder throughput. We test both schemes on a sample workload running on a SOAP based Business Intelligence software platform. Our results show that the proposed techniques can achieve an average of 10X reduction in the overall SOAP traffic. Furthermore, we show that the differentially encoded data can be further compressed by gzip to achieve a combined 15-17X reduction in the message sizes as compared to less than 5X compression obtained when using gzip by itself.