Performance Management And Monitoring Research Of The Mass Storage System

In modern society,the continuous development of information technology makes the amount of data in the industry and life growing bigger, and the requirement of the data storage capacity is also growing. The size of the storage system has been forwarding to PB-level and higher. So the mass storage system is been born under such conditions. It makes the data into stripes to organize and store them effectively, meanwhile uses the parallel file system and access protocols to provide a powerful concurrent access for the outside in the case of ensuring low overhead and latency. In addition,in order to ensure the data and system security, it also achieves the capabilities including information redundancy, load balancing and data migration.We need to know the real-time operating state of the mass storage system, so a monitoring system needed to be set up to watch its nodes and services. The mass storage system has many features of its own: first is the node expansion. The system can not anticipate when and where the new node will be added or removed. The monitoring system should find the node being absence or not automatically. Followed is the node failure. Small probability event of a single node in the ultra-large-scale mass storage system will become ubiquitous. Monitoring system need to alarm when coming to node failure. It still needs to be easy to extend. Monitoring system should present the newly added services and states of the mass storage system timely. In addition is the demand of low overhead, that is ensuring the operation of the monitoring component does not take up too many resources,and will not affect the mass storage system itself. After the in-depth consideration and comparison, we chose a monitoring model adapting to the mass storage system finally.Then against to the actual situation of the mass storage system, we find the shortcomings and deficiencies of the model itself. For a next step the model is been optimized and improved from six aspects which are scalability, reliability, manageability, security, functionality, and cost. After the perfection the model can provide more comprehensive and reliable service for the mass storage system. The node tissue is more convenient, the storage and access to the monitoring information is more secure. The function of the system is more powerful. At the same time the runtime overhead of the monitoring system is been further reduced through the optimization of the internal structure. At last we verify the validity of the model. The model improved can provide a reliable guarantee for the operation of the mass storage system.