Network Energy Consumption Analysis And Study On Energy-aware Data Transfer Algorithms

Currently,the amount of data transmitted on the Internet has exceeded the EB scale and will soon reach the ZB level each year.In order to support the massive globalization of data,network infrastructure,source nodes and destination nodes consume large amounts of electricity every moment,and the cost is difficult to estimate.About 30% of the data transmission power is consumed in the end systems(ie,sending and receiving nodes),and the rest is consumed in the network infrastructure.The research on network infrastructure power management technology has been more and more abundant,while the research work focused on energy saving of the terminal system is little,but it is the key to optimizing the energy efficiency of data transmission.The main work and results of the dissertation are:(1)This dissertation constructs a power modeling system for data transmission.According to the different access permissions of the device,the author optimizes the fine-grained power model which requires the utilization information of the four system components including CPU,memory,disk,and network card and the CPU-based power model which only needs to get the CPU usage information in the system to estimate the power consumption of the device.By comparing the estimated power consumption values with the power meter results,four different transmission tools are used to evaluate the error rates of the above models applied to different types of servers,and to verify the accuracy of the two power models.(2)This dissertation analyzes the application layer parameters for a new method of energy-aware data transmission,and describes the three application layer transmission parameters of pipelining,parallelism,and concurrency,then performs xFTP data transmission experiments on two kinds of datasets in local area network and wide area networks of two different bandwidthes.The power models were used to analyze the transmission throughput and total energy consumption at different application layer parameters levels(single parameter adjustment and combined parameter adjustment).Moreover,wide area networks of two different RTTs were targeted for three different types of datasets which were subjected to HTTP data transmission experiments,and the power model was used to analyze the transmission throughput and total energy consumption at different application layer parameter levels,confirming the validity of various parameters of the application layer.especially concurrency is most effective parameter in most cases.(3)Based on the analysis results of the impact of application layer parameters on throughput and total energy consumption,this dissertation proposes three kinds of energy-aware data transmission algorithms based on grid platform which is used to predict the optimal combination of application layer parameters for achieving optimal data transmission throughput under energy-constrained constraints and to achieve energy-consumption targets under different constraints.Algorithms include:(1)the minimum energy algorithms minimizing the overall energy consumption without focusing performance;(2)the high-throughput energy-saving algorithms aiming at maximizing throughput below low energy consumption limits;(3)the algorithms based on customers' custom requirements allowing users to define their throughput or energy consumption requirements as a part of the SLA.This dissertation analyse the comparision of algorithms and clients on energy efficiency in local area network and wide area networks,and verifies the superiority of different algorithms in different application scenarios.Experiments confirm that proper parameter combination can save a lot of energy consumption for the end system without losing much performance.With SLA-based energy-efficient data transmission algorithms,Internet service providers can minimize energy consumption during data transmission while maintaining the desired level of service quality.