Context based approach to intelligent message passing

Abstract

Being a widely used method of communication for distributed applications SOAP has a major drawback with its large message size. This is mainly because of the excessive number of tags being used on SOAP for data definition. On a frequently used web-service, some data of the message can be repeating from one message to other while only few fields of the message are changing. These characteristics of the message are totally depends on the nature of the application and the data associated with it. One cannot really predict on it at the time of the development of the web-service. The hypothesis of this thesis is that messaging in web-services can be optimized by the concept of Negotiation in Multi Agent Technology. First we identify the XML tags as the static content of the message and automatically store as a message template to use in future messages. Then we process the XML node values to further identify the static content. We define two major Agents as Cluster Agent and Text Agent. Cluster Agent and the Text Agent negotiate with each other to perform a good cluster. Then we use Genetic Algorithms to improve the cluster. The common text pattern of the cluster is identified as the static content. After we identified static content pattern of the message we keep that pattern as a pre-knowledge of the message. Then on message transferring we delete the pre-identified static content from the SOAP message being transferred. At the receiver end we again apply the same pattern to build the original SOAP message. We have tested the prototype using a sample web-service which gives a list of airports of a country when specified the country code. The test results have shown that it has been reduced 75% of the message on average by this method. Test results also show that the total time also reduced by our method and this reduction is proportional to the length of the message.