Introduction

In this tutorial, we will see how to deploy a complete power meter to monitor the power consumption of a single node or a cluster of node.

This tutorial will redirect you to other tutorial that explain how to deploy each part of the power meter (the sensor and the formula).

Prerequisites

This tutorial assumes that you have access to a mongoDB instance that is remotely accessible by all nodes you want to monitor.

CPUs of Monitored nodes must have an intel Sandy Bridge architecture or higher.

Monitor a single node

To monitor global power consumption of a single node, you can deploy all the power meter components on the same node.

First of all, deploy a HWPC sensor and connect it to the mongoDB instance : follow this tutorial

Then, deploy a RAPL formula to compute power estimation with the data collected by the sensor : follow this tutorial

Power consumption estimation will be stored in a mongodb database. See the RAPL-formula tutorial to learn how to retrieve it.

Monitor a Cluster of two node

To monitor global power consumption of a cluster you need to follow the same step as to monitor a single node.

You have to deploy one HWPC sensor on each monitored node and connect it to the mongoDB instance. You have to give a name to each sensor (with the -n parameter). The sensor name will be use by the formula to identify on which node the power consumption information was collected : follow this tutorial

When your sensors are deployed, you can start a rapl-formula to process the data collected by all the sensors. You only need one formula for all your sensors. You can deploy the formula on another node or on a node that already host a sensor. To know how to deploy a formula and retrieve power consumption data with the mongo client : follow this tutorial

When using the mongo client to retrieve the power consumption data, you can use the metadata field sensor to know the matching between data power consumption and the monitored node.

Visualisation

To visualize the power consumption of each node on a grafana dashboard follow this tutorial