Design And Implementation Of Publish/Subscribe Based Middleware

With the rapid development of Internet technology, cloud computing is a mainstream, the background of the Internet has been involved into more and more complicated distributed system. At the same time,application scope and scale are enlarging. In order to let the huge number of machines to communicate and collaborate and complete modern social computing tasks , we are in desperate need of an asynchronous,more to more communications'infrastructure, so that we can meet the Internet application system of large loose communication need.The publish/subscribe system is adapted to the dynamic large-scale distributed computing environment well and will be used widely due to its asynchronous, many-to-many and loosely-coupled communication properties.Under such background, we have done a through study on the publish/subscribe system. In addition, we developed a highly efficient algorithm for the core module event filter. By using multi-dimensional indexing ,bitwise operators, do clustering for the subscription, reduce the redundant matching and improve the system throughput.The whole system is consisted of three parts: publisher, subscriber, publish/subscribe middleware. The publishers as the information producers, pumping data to the system.The subscriber , define a subscription to the middleware , tell the system what kind of information is interesting.So consumer can get published interested news timely, reliably through the middleware. Middleware is the core part of the whole system, which is consisted of publish proxy server, subscribe proxy server, event proxy server. Events proxy server is responsible for the entire event filtering and routing, event is consisted of a set of predicates, through matching the appropriate predicates, selective distribution from the publishers'data to inform interested subscriber based on the content, which is different from traditional based on themes'matching, and bring a lot of flexibility to the system.Release proxy server is responsible for receiving the events from the producer , filter spam content, sent to publish/subscribe middleware, do matching and routing.Subscribe proxy servers, then accept subscriber's subscription information and sent to the event agency, issued to the middleware to do processing on subscription, make events matching faster.Event filter of the middleware is the core module, the efficiency is critical for the whole system response speed. By preprocessing, put subscriptions with certain common interest into a group, extract the predicates as multi-level index.According to the properties such as type category, name, operators , divide them into three levels.To the equals sign operators we construct hash table. For greater than, less than symbols according to order constraint relation,for less than a predicate index number accords with a requirement, the value of the predicate also behind accords with the requirement, can greatly reduce the numbers of matching.And we also construct subscription tree, caculate minimum number of predicates. Redeuce the matches which do not meet the conditions.Then we further improve the efficiency of system.All in all, based on the analysis and research of the previous publish/subscribe system , we design a publish/subscribe middleware framework and implement a prototype system. On the basis of the prototype ,we use different filter algorithms , compared and analyzed the result, and made some achievements.