Abstract:Datacenter capacity is growing exponentially to satisfy the increasing demand for emerging computationally-intensive applications, such as deep learning. This trend has led to concerns over datacenters' increasing energy consumption and carbon footprint. The basic prerequisite for optimizing a datacenter's energy- and carbon-efficiency is accurately monitoring and attributing energy consumption to specific users and applications. Since datacenter servers tend to be multi-tenant, i.e., they host many applications, server- and rack-level power monitoring alone does not provide insight into their resident applications' energy usage and carbon emissions. At the same time, current application-level energy monitoring and attribution techniques are intrusive: they require privileged access to servers and require coordinated support in hardware and software, which is not always possible in cloud. To address the problem, we design WattScope, a system for non-intrusively estimating the power consumption of individual applications using external measurements of a server's aggregate power usage without requiring direct access to the server's operating system or applications. Our key insight is that, based on an analysis of production traces, the power characteristics of datacenter workloads, e.g., low variability, low magnitude, and high periodicity, are highly amenable to disaggregation of a server's total power consumption into application-specific values. WattScope adapts and extends a machine learning-based technique for disaggregating building power and applies it to server- and rack-level power meter measurements in data centers. We evaluate WattScope's accuracy on a production workload and show that it yields high accuracy, e.g., often <10% normalized mean absolute error, and is thus a potentially useful tool for datacenters in externally monitoring application-level power usage.

Workload Classification & Software Energy Measurement for Efficient Scheduling on Private Cloud Platforms

A Case Study on Job Scheduling Policy for Workload Characterization and Power Efficiency

A Measurement-based Analysis of the Energy Consumption of Data Center Servers

Virtual Machine Based Energy-Efficient Data Center Architecture for Cloud Computing: A Performance Perspective.

Computing server power modeling in a data center: survey,taxonomy and performance evaluation

Workload Characterization on A Cloud Platform: an Early Experience

Infrastructure Aware Heterogeneous-Workloads Scheduling for Data Center Energy Cost Minimization

Energy-Efficient Scheduling of HPC Applications in Cloud Computing Environments

Workload-Aware Scheduling of Real-Time Jobs in Cloud Computing to Minimize Energy Consumption

Power Profile Monitoring and Tracking Evolution of System-Wide HPC Workloads

Holistic energy and failure aware workload scheduling in Cloud datacenters.

Energy efficient job scheduling with workload prediction on cloud data center

A Cloud Monitor to Reduce Energy Consumption With Constrained Optimization of Server Loads

Energy-Efficient Resource Allocation and Migration in Private Cloud Data Centre

A Task Scheduling Method For Energy-Performance Trade-Off In Clouds

Energy-Efficient Management of Data Center Resources for Cloud Computing: A Vision, Architectural Elements, and Open Challenges

Metrics based Workload Analysis Technique for IaaS Cloud

WattScope: Non-intrusive Application-level Power Disaggregation in Datacenters

Towards Energy Efficient Scheduling for Online Tasks in Cloud Data Centers Based on DVFS

Energy and Network Aware Workload Management for Geographically Distributed Data Centers

Power Management of Virtualized Cloud Computing Platform