The Prediction Method Of Information Transmission Performance For Publishing/subscription-based Distributed System

With the development of the Internet and the maturity of distributed technologies,the application of distributed systems is widely used in all walks of life.Distributed systems have emerged in the urgent need to solve large-scale system information transmission and resource scheduling optimization and distribution.However,due to the complexity and uncertainty of information interaction in distributed systems,the heterogeneity of servers,and the cross-effects of massive user scales.Distributed systems have always been a challenge in terms of information transmission performance testing.Therefore,it is of great theoretical significance and application value to select a representative research object,and extract the performance characteristics of distributed systems,and study the performance prediction of information transmission in advance,and propose a distributed information prediction method with high accuracy.Based on the research of the mainstream publishing/subscription distributed systems ' operating mechanism,this paper analyzes the business logic of publish/subscribe distributed systems,and design the general structure of publish/subscribe distributed systems,and optimizes the distribution of a distribution based on subscriptions/subscriptions.The mathematical model of the system information transmission,the specific content is as follows:Based on the fuzzy classification of existing publish/subscribe distributed systems,it is difficult to categorize and analyze common business logic.This dissertation studies the principle mechanism of publish/subscribe system and designs the basic structure of a universal distributed content distribution network.The structure clearly sorted the business logic relationships within and among the components of users,replica server service clusters,source servers,and request routing systems.For the mathematical model problem of lack of relatively comprehensive performance indicators for publish/subscribe distributed systems,the characteristics of the MMPP(2)/G/1 queuing model were studied in detail and optimized.Through the analysis,it is found that the MMPP stacking queue sys tem with heterogeneous server cascades is different from the traditional ideal server clusterin g.The MMPP(2)/G/1 overlay queuing model can hardly guarantee the full utilization of server resources,and is based on the differences in server performance.Considering adding a heterogeneous server weighting factor to avoid two iterations of the prediction error.Based on the current problems of distributed systems testing difficult to measure detailed performance indicators in real Internet environment,this dissertation chooses to test Redis,an open source distributed database system widely used on the Internet test platform Planet Lab,and to extend its rich benchmarking tools.Through the analysis of the previous business logic and the establishment of the mathematical model,the relative refinement of the measurement index can be extracted with the aid of the benchmarking tool.By comparing the relatively accurate experimental results with the predicted values of the mathematical derivation,the test results show that the optimized design of the publish/subscribe distributed systems information transmission forecasting method can accurately predict the information transmission performance of the publish/subscribe distributed system.