Design And Implementation Of A Energy Saving Storage System

With the development of information society, in order to meet the explosive growthof information storage, processing and transmission, large-scale data centers came intobeing. In a large-scale datacenters, according to the classic80/20principle, it is unwise tostore all the data in high-performance storage device. In order to allocate storage resourceproperly, there came the idea of hierarchical storage. Post studies have shown that serversin data center rarely run under100%utilization, and the server’s energy consumption isnot proportional to its utilization, the idle servers (idle CPU, idle hard disk drive etc) wastemuch energy. So it is quite necessary to design an energy-efficient hierarchical storagesystem.By analyzing the performance and energy consumption of the existing x86serverplatform, we design a hierarchical storage system based on embedded platform. Firstly,We introduced the prospects of the usage of the hierarchical storage system based onembedded platform, discussed the crucial technologies involved in this system, andanalyzed the problems we may encounter with the in the design and implementation of thesystem. Secondly, we design and implement a prototype system, which make up withthree parts: application server, storage server and management server. There are fivemodules in total in this system, which are named with system monitor module, storagemanagement module, data management module, intelligent management module and userinterface module. The system monitor module collects the data access frequencyinformation of the application server and the real-time workload of the storage server; theintelligent management module implement data migration and system schedulingaccording to those information. And then we can achieve the goal of hierarchical storageand energy saving.Finally, we verified the function of the system through experiment, tested theperformance of the system, and calculate the changes in energy consumption of the systemby scheduling the system through data migration. Our experiment results show that thissystem can dynamically migrate data based on the real-time access frequency and theworkload of the server to optimize the allocate of resources in the system. At the same time, with the help of powering on and off the servers in the data center, the system canreduce energy consumption by25%.