Not sure if this is the exact proposal text.

Ceilometer-a-la-carte: Building a custom telemetry solution

The use cases for OpenStack are vast. It can be used to launch tens of thousands of machines to meet your web service demands, create development machines in an instant for your large enterprise needs, or match the needs of your growing small business requirements. Depending on the purpose of your deployments different information is required to provide effective insight into your cloud deployment.

In this talk, we’ll show how Ceilometer is designed to cope with the diverse requirements of users’ needs. We will describe how data is modeled in Ceilometer and how these models can be used to get a better understanding of your system. We will show how you can leverage existing data and build new custom data points to track attributes specific to your environment without having to change Ceilometer. And finally we will show how you can create custom plugins to measure your systems.

Reimagining telemetry: The effects of simplification

Beginning in the Juno development cycle, Ceilometer has undergone a transformation to streamline and decouple its services in an effort to improve performance and remove complexity. The design of Ceilometer has been refactored to enable a highly scalable, componentised collection of services with greater flexibility to produce and consume telemetry data. Each service has been remodelled to perform fewer tasks but work in concert with the others to provide the same functionality as before.

In this talk, we’ll highlight the architectural changes made in Ceilometer over the past few cycles and the effect those changes have had on performance. We’ll explore the design choices made and investigate if they ultimately made any difference. By performing benchmarks, we’ll discover how changes in the storage services (SQL, MongoDB, and Gnocchi) and in the notification and polling agents has impacted the overall telemetry service. We’ll investigate performance metrics to get a deeper understanding of whether the push for simple, decoupled services has yielded performance gains.