Research On Multi-Dimensional Event Matching Technology For Massive High Frequency Information Distribution System

The content-based massive high-frequency information distribution system can achieve complete decoupling of both communication parties in space and time.Through the information distribution system,people can quickly and accurately obtain messages of interest from massive data.For information distribution systems,the event matching algorithm is the core.It is responsible for finding all subscriptions that match the event.The matching algorithm determines whether the system can work stably and efficiently in high-load scenarios.However,with the rapid increase in the number of subscriptions,most existing matching algorithms have problems such as failure to filter unmatched subscriptions in a timely manner,unreasonable division of subscription space,and imbalanced index structures.Therefore,it is of great significance to propose an efficient matching algorithm for information distribution systems,which is also the focus of scholars at home and abroad.In this paper,a code matching algorithm CEM-Tree based on ordered search tree is proposed.When inserting subscriptions,the insert branch will be selected according to the order of the attributes in the subscription.In addition,an attribute partitioning strategy is used to assist the dynamic generation of the index structure.The attribute partitioning strategy is based on matching cost analysis.Its purpose is to obtain the optimal attribute partitioning method.For the value matching problem,we convert the event value and subscription value interval into codes,determine whether the codes match the operation results,and then further divide the attribute value range into multiple cells.These cells form an isosceles triangle area in the plane coordinate system.During event matching,the overall segmentation of the isosceles triangle area can quickly filter out a large number of unmatched subscriptions,thereby shortening the matching time.In order to further improve the efficiency of event matching in the information distribution system,we propose a multidimensional event matching algorithm based on analytic geometry,which called GEM-Tree.In order to allow the index structure to adaptively adjust the distribution of subscriptions in the index structure with the continuous insertion of subscriptions,we propose a local adjustment mechanism,which contains a ranking function based on matching costs.When we choose a branch,we should choose the highest-ranking attribute branch.At the same time,in order to prevent the performance of GEM-Tree from being affected by the order of subscription insertion,we propose a global adjustment mechanism,which analyzes the distribution of inserted subscriptions to get an attribute priority,and then make a global adjustment for all inserted subscriptions according to the priority.In addition,GEM-Tree combines the plane geometry to segment the grid of the attribute value space,and the use of graphic division methods in event matching further improves the matching efficiency.In order to have a comprehensive evaluation of the performance of the matching algorithm proposed in the paper,we compared the CEM-Tree and GEM-Tree with the current state-of-the-art BE-Tree,OP-Index,and TAMA in various scenarios.On the whole,in different scenarios,the performance of GEM-Tree is better than other algorithms,and the performance of CEM-Tree is better than OP-Index and TAMA.In experiments using the number of subscriptions as the evaluation parameter,GEM-Tree was 1.6 times,6.4 times,13.3 times,and 17.2 times faster than BE-Tree,CEM-Tree,OP-Index,and TAMA,respectively.