Energy Consumption Measurement Model And Techniques For Heterogeneous Cloud Environments

With the rapid development of cloud computing technologies and applications in recent years,the number and scale of cloud data centers have rapidly increased while the problem of energy consumption in cloud data centers has become more and more serious.Therefore,energy consumption management has gradually become one of the hot issues in the field of cloud computing.The energy consumption modeling method and the corresponding energy consumption measurement mechanism are the most basic research work.In this paper,we first investigate and analyze energy consumption measurement methods and power models for servers,components,and virtual machines.Through research and analysis,we pointed out the characteristics and deficiencies of existing energy consumption models.Focusing on the heterogeneity,high scalability and high deployment cost of energy monitoring of existing data centers,the three modeling goals of data center energy modeling are expounded: high scalability,high adaptability,and high accuracy.Inspired by the deficiencies of existing models and modeling goals,this paper proposes a power-exponential CPU power model based on CPU usage.The model only uses utilization and a constant parameter to describe the CPU energy consumption with utilization.This model has good accuracy while reducing the difficulty of model training and simplifying the deployment of monitoring system.This article also investigate the relationship between disk energy consumption and performance,then a disk energy consumption model based on state partition is proposed.By using several key thresholds to distinguish between disk sequential I/O and random I/O status,make disk energy consumption measurement more accurate.Based on the above research and analysis and addressing the insufficiency of existing energy measure tools for cloud data centers,this paper designed and implemented a multicomponent based energy consumption measurement distributed tool(Distributed Energy Meter)for cloud server in heterogeneous cloud environments.DEM uses a hardware-aware CPU-based energy consumption model to select calculation methods.It also supports dynamic adjustment of model parameters,making energy consumption measurement more accurate.DEM also adopts a communication mode based on a combination of periodic push and event-triggered push,making the DEM distributed communications architecture lightweight and efficient.At the same time,DEM achieves multi-grain energy consumption measurement,including physical servers,virtual machines,containers,and processes.Finally,experiments are not only performed on a heterogeneous cluster with workload generated by PC Mark and Sysbench,but also performed on the virtual machine and containers generated by customized load script.The results show that the proposed DEM system can not only achieve energy consumption measurement of server in heterogeneous cloud environments,but also has good energy consumption measurement accuracy of cluster.