Event Details

Scheduling is one of the most important components of any Cloud deployment. Scheduling means selecting a compute layer based on user's request and taking into consideration the characteristics of the way the cloud has been setup. OpenStack compute scheduling uses a filtering and weighing mechanism to select a compute to schedule the instance on. Openstack compute scheduling works properly when resources are available to the compute locally but it can run into various complicated problems when that is not the case or when run at scale.Problems are mentioned below :

• Compute service does not report a holistic view of compute resources. This makes scheduling on those resources impossible.   
• Compute does not take into consideration resources exposed via Cinder or Neutron while scheduling.
• The scheduler lacks global view of resources (especially relevant in heterogeneous clouds).Making the scheduler aware about resource is more of a reactive process than being proactive.

In this talk we plan to cover all the above and many more interesting and scary horror stories about the broken state of OpenStack Scheduling