In the cloud environment, billing requires multiple steps and services to support it. The required steps and services are metering, rating and billing. The Ceilometer project was originally designed to provide the metering service as it gathers the data required by the billing systems that vendors would implement for their OpenStack installations. This service collects information and required data about the system and saves it in the form of samples in order to provide data about anything that can be billed.

The primary purposes of the Ceilometer are the following:

• The efficient collection of metering data.

• Collecting data by monitoring notifications sent from services or by polling the resources.

• Configuring the type of collected data to meet various requirements.

• Accessing data through the REST API.

• Expanding the framework to collect custom usage data by additional plugins.

• Producing signed and non-repudiable metering messages.

But since its inception more features have been added to Ceilometer, the most important of which is the addition of alarms that can be set to fire when monitored data streams cross user set thresholds.

Unlike the rest of OpenStack there is currently no Ceilometer dashboard component for end users. You either have to use the command line client or the RESTful web API (the command line client does offer Python bindings).

So the simplest path to follow when it comes to Ceilometer is to grab the command line tool and to get going with that.

Architecture of OpenStack Ceilometer

• API Server: provides access to metering data in the database via REST API

• Central agent: polls utilization data for resources not tied to compute nodes. There may be only one central agent running for the infrastructure.

• Compute agent: polls metering data from the compute node. Compute agents must run on compute nodes.

• Collector: monitors the message queues (data sent by the infrastructure and coming from agents). The messages are processed, converted in to metering data, signed, and put back to the queue.

• Data store: save the metering data from collectors and provide the metering data to API server.

Ceilometer meters

In the Ceilometer world, “Meters” is the term to indicate a tool that measures an aspect of resource usage.

So for any one resource there can be many meters measuring different aspects of he resource.

A meter has a name, a type and a unit of measurement associated with it.

For example:

Name: cpu Type: cumulative Unit: ns Resource ID: 0488d02d User ID: 308 Project ID: 201

There are currently three possible types of meters:

• Cumulative: an accumulating value is recorded (i.e.: the total)

• Delta: changes between values are recorded

• Gauge: only the current value at the time of reading is recorded.

Note that in older versions of Ceilometer meters were termed 'counters'.

You can also see the resource the meter is associated with, and the owning project and user.

More detail on meters, and their types and units can be found in the OpenStack documentation.

Ceilometer Meters on the NeCTAR Cloud

At NeCTAR the meters available to a given project depend upon the resources that are available to that project.

So for a project that has volumes, these are the meters that would typically be seen that relate to the guest machines at this moment in time:

The following are the meters typically seen on a project that relate to Glance:

The following are the meters typically seen on a project that relate to Swift:

The following are the meters typically seen on a project that relate to Cinder:

A complete list of meters, and their types and units can be found in the OpenStack documentation.