An intelligent power consumption model for virtual machines under CPU-intensive workload in cloud environment

Abstract

Cloud computing has gained enormous popularity by providing high availability and scalability as well as on-demand services. However, with the continuous rise of energy consumption cost, the virtualized environment of cloud data centers poses a challenge to today's power monitoring system. Software-based power monitoring is gaining prevalence since power models can work precisely by exploiting soft computing methodologies like genetic programming and swarm intelligence for model optimization. However, traditional power models barely consider virtualization and have drawbacks like high error rate, low feasibility as well as insufficient scalability. In this paper, we first analyze the power signatures of virtual machines in different configurations through experiments. Then we propose a virtual machine (VM) power model, named CAM, which is able to adapt to the reconfiguration of VMs and provide accurate power estimating under CPU-intensive workload. We also propose two training methodologies corresponding to two typical situations for model training. CAM can estimate the power of a single VM as well as a physical server hosting several heterogeneous VMs. We exploited public Linux benchmarks to evaluate CAM .The experimental results show that CAM produced very small errors in power estimating for both VMs (4.26 % on average) and the host server (0.88 % on average).

A Power Consumption Model for Cloud Servers Based on Elman Neural Network

“…Previous studies mainly used regression-based methods to build power consumption models. The most widely adopted approach is linear regression because of its good interpretability and simplicity in training [9]. For example, Hsu and Poole [10] investigated a number of regression-based power consumption models which are all functions of CPU utilization.…”

A Workflow Scheduling Deadline-Based Heuristic for Energy Optimization in Cloud

“…We assume that this cost is directly correlated to the data center's power consumption. It has already been shown [11], [12] that the CPU's energy consumption is not a linear function defined by the load and the running time. Therefore, reducing the number of nodes for a longer period of time is not equivalent to using more nodes for a shorter period.…”

“…Therefore, reducing the number of nodes for a longer period of time is not equivalent to using more nodes for a shorter period. To properly model the cost and therefore the gain of our solution, the consumption of a node is defined accordingly to [11] and as follows:…”

“…Intuitively, when trying to reduce the makespan of a workflow, it is necessary to use a large number of machines, which is costly for the cloud provider. Indeed, since energy consumption cannot be represented by a linear function [11], [12] defined by the number of machines and the running time, the use of a large number of machines for a very short time is more consuming than the use of a reduced number of machines for a slightly longer time interval. Conversely, if users can extend their deadlines instead of looking for the best possible makespan, a better energy optimization (i.e., number of machines used) can be achieved, resulting in a cost reduction for the Cloud provider.…”

A CO2 Emissions Accounting Framework with Market-based Incentives for Cloud Infrastructures

“…They are usually based on counters (hardware or software) in order to monitor the resource usage. Their accuracy thus depends which resources are selected, how they are monitored and which formulas are used to estimate the VM power consumption from the monitoring data, such as linear regression (Kim et al, 2011) (Wu et al, 2016), polynomial regression (Xiao et al, 2013), machine learning (Yang et al, 2014) or tree regression based approach (Gu et al, 2015). In these studies, estimation errors typically fluctuate from 2 to 5%.…”