Open Stack

I would like to start discussion regarding user experience when certain
level of network QoS is expected to be applied on VM ports. As you may know
basic networking QoS support was introduced during Liberty Release
following spec, Ref [1]
As it was discussed during last networking-QoS meeting, Ref [2], nova team
drives to the approach where neutron port is created with all required
settings and then VM is created with pre-created port  and not with
requested network. While this approach serves decoupling and separation of
compute and networking concerns, it will require smarter Client
orchestration and  we may loose some functionality we have today. One of
the usage scenarios that currently supported, is that Cloud Provider may
associate certain requirements with nova flavors. Once Tenant requests VM
for this flavor, nova (nova-scheduler) will make sure to fulfill the
requirements. Possible way to make this work for networking -qos is to set :
 nova-manage flavor set_key --name m1.small --key quota:vif_qos_policy
--value <the-policy-id>

With current VM creation workflow  this will require nova to request
neutron to create port and apply qos policy with specified policy_id. This
will require changes on nova side.
I am not sure how to support the above user scenario with pre-created port
approach.

I would like to ask your opinion regarding the direction  for QoS in
particular, but the question is general for nova-neutron integration.
Should explicitly  decoupled networking/compute approach replace the
current way that nova delegates networking requirements to neutron.

BR,
Irena


[1]
http://specs.openstack.org/openstack/neutron-specs/specs/liberty/qos-api-extension.html
[2]
http://eavesdrop.openstack.org/meetings/neutron_qos/2015/neutron_qos.2015-09-16-14.02.log.html
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20150924/88d05b77/attachment.html>