OpenStack Liberty Design Summit

Expressing (implicitly or explicitly) a rich set of computational requirements when starting an instance is a key factor to the success of enterprise-grade deployments that require high-profile resources. Making sure such Service Level Agreements for Quality of Service requirements can be met either now or in the future is critical to this success.


In this talk, we will cover the existing possibilities that OpenStack uses to specify a specific set of computational resources. We will examine how to influence the placement decision by going under the hood of the existing Nova scheduler. We will discuss some issues with the current design and present current efforts for splitting the Nova scheduler into a separate project called Gantt. Finally, we will address how other OpenStack projects, like Cinder and Neutron, can take advantage of a common scheduler.