Performance analysis is a critical requirement for developers who desire to integrate new features (regression testing) and track bottlenecks for specific OpenStack deployments and settings. Since no one can handle to deploy and benchmark OpenStack manually, we must provide i) a simple way to express customizable OpenStack deployments with reusability in mind; ii) automated tools to deploy and perform the benchmarks.
In this talk we present EnOS, a holistic framework to conduct easy and reproducible evaluations of different OpenStack deployments. Given a highly customizable configuration, EnOS can:
- Provision the required machines for various testbeds (Vagrant, Chameleon, OpenStack, G5k)
- Deploy the desired OpenStack services with specific settings
- Deploy performance analyzing tools
- Run data plane (shaker) and control plane (rally) benchmarks
- Analyze and backup the results
This talk will review some use-cases representing joint works between the Performance and FEMDC WGs.