Orange, as many other network operators, has multiple PoP (Point of Presence) sites geographically distributed where network functions are delivered to its clients. In the NFV context, these edge sites may host VNFs that are run on compute nodes remotely connected to a centralized control plane through a Wide Area Network which exposes traffic to high latency, possible packet loss, and other network hazards.
In the presentation, we will show our performance evaluation of Edge OpenStack deployments, with few hundreds of compute nodes connected to the control plane through a Wide Area Network. First, we focus on the performance of the message bus used for OpenStack's interprocess communication, and the suitability of the underlying technology (RabbitMQ, Qpid Dispatch Router) for edge scenarios. Second, we explore the survivability of the message bus to network disasters. We conclude by pointing out the other challenges that should be addressed in such scenarios besides the message bus.
What is the performance of the message bus used in OpenStack's interprocess communication and the suitability of the underlying technology (RabbitMQ, Qpid Dispatch Router) in edge computing use case.
