Research And Design Of Linux Server Performance Monitor And Analysis System

With the rapid development of Internet service, the servers of Internet-related enterprise are bearing more and more pressure, and always have expansion needs. To ensure stable operation of the servers, monitoring the servers’performance has a very practical significance for those enterprises. However, even though the current technique of performance monitor can provide real-time performance data of servers and generate alarm when the performance occurs bottleneck, but it lacks of the storage and query of historical data and analysis of performance data and some other aspects of features. Based on the existed monitoring technology, this thesis designs and implements a set of Linux server performance monitoring and analysis system.Firstly, this thesis introduces a series of traditional server performance monitoring technology and data analysis techniques. And then we describe in detail the design and implementation of the module of performance data acquisition and transmission, including the comparison of the performance of different kinds of long-distance data transmission methods and the auto update of the module version. To deal with the massive performance data produced, this thesis uses Hadoop data processing framework to establish a set of performance data storage and query system, the system uses HDFS(Hadoop Distribute File System) to store performance data in binary mode, while taking advantage of the MapReduce framework to write daemon to achieve concurrent queries of performance data. After obtaining historical performance data using this system, we define the anomalies by observing the changing curves of all kinds of performance data, including the burst anomaly, the distribution anomaly and the memory leak anomaly. Then we come up with some corresponding methods, for example, dynamic threshold method and function model method to detect these anomalies according to their features, and the reasons that cause these anomalies are also analyzed in a brief way. Finally we summarize and discuss the prospect of server performance monitoring.