Online Power Consumption Estimation for Functions in Cloud Applications

Abstract

The growth of cloud services leads to more and more data centers that are increasingly larger and consume considerable amounts of power. To increase energy efficiency, informed decisions on workload placement and provisioning are essential. Micro-services and the upcoming serverless platforms with more granular deployment options exacerbate this problem. For this reason, knowing the power consumption of the deployed application becomes crucial, providing the necessary information for autonomous decision making. However, the actual power draw of a server running a specific application under load is not available without specialized measurement equipment or power consumption models. Yet, granularity is often only down to machine level and not application level. In this paper, we propose a monitoring and modeling approach to estimate power consumption on an application function level. The model uses performance counters that are allocated to specific functions to assess their impact on the total power consumption. Hence our model applies to a large variety of servers and for micro-service and serverless workloads. Our model uses an additional correction to minimize falsely allocated performance counters and increase accuracy. We validate the proposed approach on real hardware with a dedicated benchmarking application. The evaluation shows that our approach can be used to monitor application power consumption down to the function level with high accuracy for reliable workload provisioning and placement decisions.