Online Health Status Assessment For Publish/Subscribe Distributed Systems

Data distribution service(DDS)based on publish/subscribe model is a distributed real-time middleware communication specification proposed by the OMG organization.It provides an efficient communication mechanism and supports the data-centric distributed application scenarios.Therefore,it has been widely used in national defense,civil aviation and other key areas.In these areas,the publish/subscribe distributed systems become very important because they carry critical applications.So the online monitoring systems are necessary to Ensure that the publish/subscribe systems work properply.However,existing monitoring software can only gather running status information,system maintenance personnel can only determine whether the system is normal based on their experience through a comprehensive analysis of the status information obtained.Therefore,the research on theories and methods of comprehensive online evaluation of the overall health status of a system is now in urgent need.In this thesis,an online health status assessment solution is designed by proposing an index system of the health status assessment of publish/subscribe distributed systems and the corresponding assessment method,based on the analysis of the characteristics for publish/subscribe distributed system,which can real-time collects key operating status and online comprehensive assessment for publish/subscribe distributed system,,allowing users to monitor the operation of the system and initiate an alert by observing the dynamic trends of the assessment results.Main contributions are listed as follows:1.Index systems of health status assessment for publish/subscribe distributed system is proposed,which adopted a hierarchical structure and is able to decompose the complex issue of comprehensive assessment into multiple indexes at different levels and adjusted its overall hierarchy according to the number of assessment objectives detected in the system.2.Method of Health status assessment for publish/subscribe distributed system is proposed.Based on the hierarchical structure of the index system,the expert experience and the Pagerank algorithm are employed to work out the weight of each index in the method,in addition,Z-Score and linear transformation methods are used to preprocess the collected status data,finally calculate the system health status value.3.A prototype system of online monitoring and condition assessment for publish/subscribe distributed system is designed and implemented,based on the aforementioned online monitoring status assessment program,the prototype system collects the operating status of the publish/subscribe distributed system online by using Flume and Spark Streaming is used to calculate the system health value,the trend of system health value is displayed on the interface.Results of the functional and performance tests of the prototype system indicates that it can accurately assess the health status of the system and meet the online assessment requirements.