OpenStack Neutron & OpenVSwitch: Speed your provisioning and scale much better with VxLAN

If you want to be or stay as a market leader, you have to be humble enough to open your mind to more options than those that your team has been developing. The fact is, there will be always a better way than yours, and if it doesn’t exist yet, don’t worry… it will come soon.

Last week, I had the pleasure to attend to our so awaited classic event @KloudCamp in Mexico and meet some amazing people. This @KloudCamp has got more sponsors than ever which enriches this regional event even more – I sacrificed some amazing sponsor’s presentations to dive into other matters and chats, but I don’t regret, as a matter of fact, I would have wished having more time to chat and learn more things from those so relevant guests.

I have to recognize that Nadia Mansour, the regional Arista Networks representative, was more than persuasive to take me from somewhere @KloudCamp to chat with Anoop, and I must admit that our conversation at the end it’s worth every minute. Anoop started to draw some objects on a piece of paper in regard to what Arista could make to improve the integration of OpenVSwitch with OpenStack. Honestly, I didn’t see so much value at the beginning about changing an architecture that has got so much time from us to implement. An architecture that fulfills more than enough our local market demands and its understanding – we firmly believe we are the only provider in LATAM with advanced SDN functionalities in our cloud offering –

Anoop’s suggestion was to replace GRE tunneling with VxLAN, which means the encapsulation process will be processed by switches removing it from the tunnel bridges that OVS usually defines into the Nova and Neutron nodes. I have to say that I’ve found this suggestion more than useful. It will give us more than one advantage:

  1. This new approach will free up some compute resources that currently we have assigned to encapsulate traffic through GRE among NOVA and NEUTRON nodes. I don’t know exactly how much cost savings we will get, but I believe It will be important depending on as much your cloud scales.
  2. It will be much simpler to interconnect cloud stacks installed on different datacenter’s cluster or locations. Using VxLAN over a smarter network topology – VxLAN works on top of a layer3 configuration – it will be so much easier to provision and operate them than define exclusive traditional VLANs between these places and make it scalable over a Layer-2 topology throughout a typical configuration of aggregation+core switching. In other words, VxLAN helps us to have a more transparent and faster way to establish cloud interconnections. In fact, you don’t need to touch your current core switching infrastructure at all to work under this new concept.

Well, thanks for reading. See you next!

2 thoughts on “OpenStack Neutron & OpenVSwitch: Speed your provisioning and scale much better with VxLAN

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s