Erez Cohen

Can the Open vSwitch (OVS) bottleneck be resolved? Erez Cohen, Mellanox

OpenStack practitioners who have deployed OpenStack cloud at scale would frown when they hear the mention of Open Virtual Switch (OVS), which has been a bottleneck for cloud network performance and scalability. As emerging technologies such as NFV and other performance-sensitive cloud deployments keep pushing for higher data forwarding performance across the network infrastructure, it becomes critical to improve OVS performance without compromising flexibility, network programmability, and cost.

In this session, we will present a novel way that Mellanox has developed to offload the entire OVS dataplane onto the embedded switch (eSwitch) implemented in the server Network Interface Card (NIC)’s hardware. This approach can not only boost server I/O performance to near line-rate, but also doing so at a fraction of the CPU load needed by existing OVS implementation. Ultimately it maximizes the effective bandwidth that the applications can use to communicate with each other or fetch data from storage, and enhances the efficiency of the cloud. Accelerated Switching And Packet Processing (ASAP2) Direct works seamlessly within the framework of Software Defined Networking, and allow SDN controller to configure and update flows onto OVS the same way as before so that network programmability remains intact.

In this session, we will introduce OVS offload with ASAP2 Direct technology and show how it is seamlessly integrate with OpenStack.