Study On Computer Cluster Node State Measurement And Optimization Of Energy Saving

With the deepening of social informatization, the size and number of computer clusters are increasing rapidly. So the coming problems are: 1) It’s hard to get the whole picture of the cluster state, which could lead to high temperature of local clusters, and affects safe and stable running of the clusters; 2) It needs to turn down the temperature of refrigeration air-conditioning to prevent high temperature of the clusters, but it will also lead to overall energy consumption of the computer clusters heavily. In recent years, many solutions of cluster monitoring and energy-saving optimization have emerged, but they all have varying degrees of defects, such as the problems of less collection points and lo w sampling frequency of the cluster monitoring system. In energy-saving optimization, only optimization and energy consumption of the single node in the clusters is considered instead of considering the overall energy consumption of the clusters, and so on.According to the two problems above, an intelligent fine-grained monitoring system design scheme of monitoring temperature and wind speed is proposed, as well as an energy-saving optimization scheme of computer clusters based on relationship between environmental data and load. These schemes can achieve real-time monitoring of environment data and comprehensive energy-saving of the computer clusters. The following work is done in this paper:1) This paper presents a software/hardware design and realization of monitor ing system of computer cluster environment data. Upper/lower computer data transmis s io n structure is used for environmental data monitoring system. Among them, the STC89C58 RD is selected as the environment data acquisition module in controller chips, DS18B20 as temperature collecting sensor, MS5611 as indirect acquisition sensor of environment wind speed, and nRF24L01 as send chips of wireless transmission of environmental data in lower computer; while in upper computer, STC89C58 RD is chosen as controller chips for receiving environmental data, as a consumer of data control, nRF24L01 as receiver chip for wireless environmental data receiver and transmitting data to the management node through the serial port; In order to communicate with Linux systems, a Linux communication driver for serial port and data display as well as store programs on Linux system is developed.2) A model for predicting air outlet temperature of cluster nodes in energy-saving optimization of the clusters is proposed. The model can be got by building the model for predicting air outlet temperature and taking advantage of environmental data, physical resources and the historical data of the virtual machine resource of the clusters. This model can be used to calculate the heat contribution of virtual machines on the physical node.3) An energy-saving optimization algorithm based on relationship between environmental data and load in computer clusters is presented. This paper gives the energy-saving optimization algorithm based on relation between temperature and load, and describes the design and implementation process of initialization algorithm of resource deployment and thermal balancing migration algorithm in running process. Energy-saving optimization algorithm is further studied based on relationship between temperature, wind speed and load in the end.Comprehensive monitoring and optimization of energy saving is achieved in the computer cluster monitoring and energy-saving optimization scheme presented in this paper. And it also helps to reduce the workload of cluster management and the overall energy consumption of the clusters.