VXLAN is hot. We constantly hear about VXLAN at conferences, in product announcements, blog posts, and most importantly, we hear about it from our customers.
VXLAN exciting technology that’s been integrated into a number of product offerings from networking and cloud vendors. OpenStack® supports VXLAN via a set of Neutron plugins, and Metacloud OpenStack® has supported VXLAN for a few releases already.
One of the challenges with deploying and scaling VXLAN has been the MAC-to-VTEP learning and BUM (Broadcast, Unknown Unicast, Multicast) packet flooding. The VXLAN spec uses a simple multicast solution to solve this problem. Multicast has its own set of scaling challenges, and reliable multicast routing between network segments isn’t always available. The majority of vendors who have VXLAN support have attempted to solve this problem by implementing their own form of learning and flooding. Some of these solutions work well, but all of them require you to operate in a homogenous network environment or pay expensive per CPU or per VM licensing fees.
Metacloud, in partnership with our friends at Cumulus Networks®, have been working together on a solution to these problems for the past year. Starting today, VXFLD is open source and freely available on GitHub (https://github.com/CumulusNetworks/vxfld) under GPLv2.
Our goal was to solve these challenges for our clients who wanted a VXLAN solution that didn’t cost them a fortune or lock them into a specific vendor configuration. To that end, we have created a python-based service called VXFLD that enables unicast based MAC-to-VTEP learning and BUM packet flooding. The service is written in python and is hardware-agnostic. It is designed to run on any operating system with a working python implementation, and can handle VXLAN traffic from any operating system or hardware platform that adheres to the IETF Internet-Draft for VXLAN (https://datatracker.ietf.org/doc/draft-mahalingam-dutt-dcops-vxlan/).
This is only the beginning for VXFLD. Over the next several months we have plans to introduce a head-end replication mode for better scaling, and a proxy mode that will allow for tiering of service nodes and flooding. We look forward to feedback and contributions from the community.
At Metacloud we are very excited about the fact that there is now a free, as in beer, and a free, as in speech, VXLAN solution. This is an important step for OpenStack networking, and we hope you’ll get involved in the new open source VXFLD project on GitHub. Let the Open VXLAN networking revolution begin!
Read more here:: Cumulus Networks