Research On Performance Monitoring And Dynamic Scheduling Technology In Distributed MCPTT System

MCPTT(Mission Critical Push to Talk)system is a digital trunking communication system based on LTE technology,which is used in command and dispatch of mission critical scenarios such as government,public security and transportation.Limited by the server capacity,with the increase of user capacity,the centralized MCPTT system has problems such as concurrent service collision and large service response latency.The distributed MCPTT system horizontally expands the components of MCPTT system to solve the capacity limitation of server and increase user capacity.Affected by load balancing and node scheduling mechanisms,the concurrent performance of distributed MCPTT system services is poor.The performance monitoring and dynamic scheduling technology of distributed system can improve the concurrent performance of system services.Therefore,the research on performance monitoring and dynamic scheduling technology in distributed MCPTT system is of great significance.On the basis of reviewing the research status of domestic and foreign distributed system,performance monitoring technology and dynamic scheduling technology,this paper studies and realizes the performance monitoring and dynamic scheduling technology of distributed MCPTT system.The main work is as follows:1.Based on siding time window real-time monitoring principle,the performance monitoring mechanism of distributed MCPTT system is designed and low-latency node running state monitoring is realized which can provide data basis for request load balancing and cluster auto scaling technologies of the system;2.A dynamic load balancing algorithm based on load feedback is proposed.Based on the large request difference of MCPTT system and the disadvantage that the traditional algorithms didn’t consider user and request differences,this algorithm dynamically adjusts node allocation strategy according to node real-time monitoring data,which resolves the request allocation problem in system hierarchical calling and improves the cluster load balancing degree of the system;3.A node scheduling algorithm based on load forecast is proposed.Based on slow node start-up of MCPTT system and response hysteresis of threshold-triggered auto scaling mechanism,load forecast and node preloading mechanisms are introduced to realize efficient node dynamic scheduling,solve performance degradation and resource waste caused by system load fluctuations and improve the average resource utilization of the system;4.Based on Linux operating system,using C++ and Java languages,modules such as real-time performance monitoring,request load balancing and cluster auto scaling are developed,which realize real-time performance monitoring,cluster dynamic load balancing and cluster auto scaling of distributed MCPTT system;5.The verification environment of the performance monitoring and dynamic scheduling technology of distributed MCPTT system is built and the verification content of the performance monitoring and dynamic scheduling technology is designed,which are used to verify the function including user login,user affiliation,user call and performance monitoring and the performance of system including cluster load balancing degree,average resource utilization and request reply latency.The verification result shows that the implemented performance monitoring and dynamic scheduling technology in distributed MCPTT system meets the function and performance requirements of the system.