This is the first public demo of a new major feature guaranteeing bandwidth on network ports.
With this feature a cloud admin can create a QoS policy with a guaranteed minimum bandwidth rule. The user can attach that QoS policy to a Neutron port. So instances will be scheduled properly by Nova where the required bandwidth is available. (On the closest physical network interface.) This means that other instances sharing the same compute host cannot hurt the network performance of such an instance.
How did we get here? For SR-IOV ports we had partial support already since Newton. At that time we could not yet avoid accidental violations by co-locating instances needing more than what was available on a compute host. Since then we introduced the Placement service to model the total/used/free resources across the whole cloud. We also worked to integrate Neutron and Nova with Placement.
- Miguel Lavalle
- Balazs Gibizer
- Slawomir Kaplonski
- Bence Romsics
On this session you will learn about:
- The architecture of this feature across Placement, Nova and Neutron.
- As a cloud admin how to configure your cloud to guarantee bandwidth.
- As an end user how to get guaranteed bandwidth.
- As a developer you will get some insight how to work with cross-project, multi-release features.
This feature is one of the first users of the Placement service outside of the Nova project. To help the understanding you will also get a quick Placement intro. And a live demo of course.