Design And Implementation Of Cloud Computing Platforms Monitoring System

Currently, Cloud Computing is one of the hottest research subjects. Monitoring for Cloud Computing, which is one of important parts of Cloud Computing, is gradually becoming the focus of attention.This thesis designs and implements monitoring system for Cloud Computing depending on national public welfare ocean project after researching current monitoring systems.Firstly, this thesis introduces and analyses relative monitoring techniques, mainly focus on current mainstream monitoring systems for Cloud Computing.According to analysis of structure and features of current monitoring systems, the thesis recommends a monitoring system architecturet applicable to systems corded by Hadoop.Secondly, this thesis explains and introduces the design of this recommended system architecture in datails,this system is divided into Client level and Server level.Client level calls Server level’s services,Server level complements all the system’s functions. Then Server level is divided into agent and monitoring service machine two parts.Agent gets experimental data, monitoring service machine collects, stores, processes and displays experimental data respectively by collector module,storage module,processing module and display module.According to practical problems, this thesis mentions time gradient,first acquisition, data dilution and other strategies. Time gradient resolves the problem of large amount of monitoring data occuping bandwidth in short time. First acquisition resolves the problem of static information is acquired many times.Data dilution resolves the problem of storage system’s limited capacity.At the end of the designing stage, this thesis designs communication protocol of this system.The system complements transferring of experimental data and monitoring orders,meanwhile, implements effectively communication between agent and monitoring service machine under the protocol.At last, this thesis uses java programming language and is implemented by socket,multiple-thread,serialization and deserialization techniques.This thesis tests the monitoring system’s real-time monitoring and hadoop log’s analyzing function,the result of test shows this system realize the two functions mentioned above fairly well.What’s more,this system’s performance is rather well to satisfy current application environment after analyzing of this system’s performance evaluation.