Today OpenStack is able to track the state of hypervisors hardware with nova compute running on each hypervisor node. With state of the hardware known, OpenStack can do a better job at finding out the Hypervisor that matches the resource needed to host a Virtual Machine.
Although Neutron agents collect some state data for the Network Nodes, Neutron has no knowledge of the state of the physical networking devices.
In this presentation, we will discuss the following ideas to leverage network health monitoring for your OpenStack workloads:
- Nova to Neutron interaction for physical network details - we will discuss the APIs that allow operators to integrate the physical network information for use in workload placement
- Using Ceilometer to collect physical network stats - we will discuss how health of physical network can be modeled to provide operational insight
- Network aware Nova Scheduler - a Nova scheduler that considers the network conditions while selecting a hypervisor for the workload.
- Just like physical server infrastructure, the physical network infrastructure plays a critical role in optimal workload placement
- How Neutron can be extended to provide abstract APIs about physical network
- How Ceilometer and Nova can consume these APIs to (a) gather network health monitoring data (b) provide useful insight about network health and (c) place the workload using the network health insight.