Research Of Storage System Optimization Based On Thermal Estimation

As the developing of the information technology, the amount of data growsexponentially, it is very essential to optimize the performance and to lower the powerconsumption of the hard disk. It is also the response to the call of the contribution to theenergy saving. If you want to improve the performance of the disk without considering thepower consumption, it will lead to waste of energy. Meanwhile, simply to reduce thepower consumption of the hard disk just ignoring the performance also appears to havelittle significance. A Large number of experimental datas show that the temperature ofstorage system is the key factor, so temperature is the condition that can not be justignored. Therefore, we must stand on the point of view that not only improve performancebut also reduce consumption, It is possible to find a balance between performance andpower consumption, so as to save energy but not weak the performance.This paper presents a storage system optimization system, through the temperaturecontrol to the hard disk in order to get the purpose of improving performance and reducingpower consumption. First introduced the data acquisition unit, which is divided into fourmodules：device management, data storage management, load occurs and the commandreceiving module. The temperature prediction unit takes advantage of conservation ofenergy together with the use of thermodynamic formula to model the hard disktemperature. Then, put the historical thermal data collected by the acquisition unit as inputto the RLS algorithm to predict the parameter of the temperature module. It tested to havea high accuracy.Secondly, when the trend of the temperature found to be too high, there is a need totake a control method to prevent the disk from continuing to rise. There are two thermalthresholds. When the temperature is going to be higher than the first threshold, transfer theload of the corresponding disk, if it doesn’t work obviously, and the temperature continuesto higher, spin down the disk. The design of a virtual device layer is to make request transparent to upper. The virtual device contains two parts, one is capacity management,which is in charge of manage the storage pool; the other is address remapping, which cantake the request to another disk.Finally, do a number of related system testing. First, thermal prediction accuracy wastested, and the results showed that the prediction function has a high accuracy, and thethermal control makes some improvement to the overall energy consumption of thesystem.