Research On Matching Algorithm Of Content-based Publish And Subscribe System

The continuous expansion of computer network coverage and the rapid development of Internet technology have greatly changed the scale and complexity of distributed systems.The changes in the scale and complexity of the system impose more stringent requirements on the dynamics of the communication model and the loose coupling of the system.The traditional synchronous communication model has some characteristics,such as tight coupling,static,etc.These characteristics make it a certain threat in asynchronous transmission,multi-point communication,large-scale transmission and so on.The publish/subscribe system is a kind of middleware system,which has the characteristics of asynchronous transmission,multi-point communication and loose coupling,so that the two parties can complete decoupling in time,space and control flow.The content-based publish/subscribe system has gradually become a hot topic for people to study because it has the following features: rich expression capabilities,a more flexible subscription language,and the ability to perform retrieval operations on event information with fine granularity.Although the content-based publish/subscribe system is very flexible and expressive,it also brings great challenges: there are more subscriptions in the system,and the system matches the events based on the subscription conditions.Therefore,an efficient matching algorithm is needed to improve the matching efficiency.This article has conducted in-depth research on matching problems.The main research results are as follows:(1).An event matching algorithm named AVDDM is proposed.This algorithm constructs an index structure based on multiple attribute chains.Each chain is divided into several groups by attribute value domain,and the subscription classification is stored in the corresponding chain.In the corresponding group.The division of attribute domains can effectively reduce the number of subscriptions that participate in matching in the subscription,and improve the matching efficiency to some extent.Theoretical analysis and experimental data show that AVDDM has certain advantages in matching time,insertion time and memory overhead.(2).A new index structure Dh-Tree is designed for the matching problem of subscriptions with several common atomic constraints.Dh-Tree selects multiple public attributes as index attributes,divides the attribute range into several units,and links these units together to form a hash tree.Dh-Tree is essentially a hash table consisting of multiple chains.When the events match,many unmatched subscriptions can be filtered out,matching only those subscriptions with a higher probability of matching.Theoretical analysis and experimental data show that Dh-Tree's matching speed average is five times that of similar algorithms.