Research On The System-level Of Temperature And Power Consumption Equilibrium Model

With the soaring of the processor’s integration and performance in the current, the products embedded with multi-core processors have been more widely used in daily life. People have rising demand for the product performance, at the same time they put forward higher requirements on the heat and power consumption. While the processor acts as the system’s main heat source and power consumption unit, therefore it becomes a hotspot of current research.At the first of the paper, the current researches about temperature and power consumption are presented on different classification criteria. Based on the detailed analysis of them, the temperature and power equilibrium model of multi-core processors are investigated from the perspective of Linux system which is widely used in the operating system. The thermal characteristics of the cores and tasks are calculated by using the hardware performance counter, according to the calculation results, the tasks are allocated or migrated between every process core, meanwhile the improvement measures for the system scheduling algorithm are proposed. In the selection of task allocation strategy, the allocation strategy combined the coldest core distribution and the historical trend of temperature strategy is proposed. In the thread migration strategy, the coldest process migration is proposed, and in the scheduling algorithm, put forward the improvement TBCFS algorithm for Linux’s default scheduler CFS in order to dynamically adjust the ready queue in the processes. The purpose of the research for the model is to make the temperature of processors distribution more uniform and the power consumption growth more gentle through improved the operating system scheduling strategy without reducing the processor performance.The last part of the experiment, this paper demonstrated the task allocation strategy and scheduling algorithm of TBCFS using the HotSpot temperature and power consumption model. The simulation results show that the former can make the temperature distribution more uniform between the processor cores, and the latter can effectively reduce the peak temperature of the unit processor. Therefore the temperature was improved which has a positive impact on the power control.