Distributed Publish And Subscribe System Based On Spatio-textual Data Stream

With the rapid development and widespread application of the global positioning system and the development of social media technology,the network is closely related to people's daily lives.A large amount of social information containing spatial location data is published on the network,and interested people use social software Capture on the top to facilitate daily life.Such social activities have very strong spatial limitations and timeliness of information.People's living space is limited to a certain area,and the information in this area can more strongly affect the people living in this area.Similarly,information is time-sensitive,and the timeliness of receiving information affects people's judgments,so if timely access to important information can greatly facilitate people's lives.This is the publish and subscribe system studied in this article.The publisher publishes the information to the network,and the people who are subscribed to download it in time,which can greatly facilitate people's daily life.The traditional publish and subscribe system ignores the study of timeliness,and the use of spatial information is not sufficient.The publish-subscribe system in this paper,based on a topology model,proposes an instant publish-subscribe algorithm that can be applied on the distributed system,and then proposes to apply self-organizing incremental learning neural network(soinn)to the original algorithm.Optimization algorithm on the Internet,and proposed a load balancing strategy of the algorithm on distributed systems,and finally,a hash optimization strategy of the algorithm was als o proposed.The specific content includes the following aspects:(1)Instant publish and subscribe algorithm: At present,most of these system algorithms are batch processing algorithms.This paper proposes an instant publish and subscribe algorithm on a distributed system,which coordinates attributes and spatial information,and can be updated in time with the information flow Cluster data.(2)Application of self-organizing incremental learning neural network: the characteristics of soinn's self-organizing and incremental learning,after fully learning the self-organizing incremental learning neural network,apply it to the publish-subscribe algorithm of this article.Optimized the time and clustering effect.(3)Load balancing strategy: In a distributed system,data transmission occupies a lot of time.In this paper,this problem is considered,and some optimization strategies are carried out for data exchange between clustering categories to minimize the difference in data.The transmission frequency and size of the transmission content between nodes.(4)Hash strategy: In order to speed up the comparison between the published data and the subscribed data,in addition to the optimization of the appealing algorithm,this article also proposes a suitable hash strategy in the comparison step to reduce the number of attribute comparisons.Thereby improving efficiency.(5)Experimental verification: This article uses real data to verify the proposed algorithm and optimization strategy on a self-built cluster.The experiment shows that the publish and subscribe algorithm of this article meets the characteristics of instant and self-organization on a distributed system..And there is an improvement in time.