The Design And Implementation Of The Linux Power Management System Based On Resource Monitoring

With the popularity of computers, energy saving in computers has become notonly an important part in constructing our saving society, but also critical needs of theusers. Power management is one of the effective energy saving method in computers.In this paper, we mainly focus on one important direction which is dynamicfrequency scaling technique. Dynamic frequency scaling technique has been provento be an efficient way to achieve low power consumption. The essence of dynamicfrequency scaling technology is the CPU frequency dynamically adjusted to providethe right speed of the electric circuit to process the load of the system and meet thetotal computation time and all of the constraints, thereby reducing energyconsumption.First,we introduces low-power,power management,dynamic frequency scalingtechnology,and four algorithms of dynamic frequency scaling.We studied resourcemonitoring technology for XEN, proposed a resource monitoring method for thesystem：using Xentrace, Xenstore, read data in VM to get the information of CPU,memory, IO in monitoring the virtual machines. In addition, This paper studies theprediction algorithm such as regression analysis and proposed a new predictionalgorithms.Secondly, based on the proposed prediction algorithm,we design and implementthe Linux power management system based on resource monitoring finally. Thesystem includes three modules: resource monitoring module, predictive analysismodule, the frequency adjustment module.We introduce the design of modulefunctions and information collection in resource monitoring module. We introducethe design and implementation of this module in predictive analysis module. We usethe Linux cpufreq mechanisms to achieve frequency adjusting module, introducingcpufreq system, internal mechanisms, and the way to set the CPU frequency.Finally, we test the system. The experiments is divded into two parts:thevalidation of predictive analysis and testing enery saving effect.The experimentresults show that the proposed algorithm can well fit the CPU,memory,IO utilization,and power management system implemented in this article can effectively reduceenergy consumption effectively.