Hi

In addition to what we said at the meeting, I’d like to mention that what the 
draft is proposing seems to be a subset of what SD-WAN vendors are doing. I 
apologize in advance for using the terms NSF and SD-WAN gateway interchangeably.

For those not familiar, Wikipedia has a so-so article about it: 
https://en.wikipedia.org/wiki/SD-WAN

In an SD-WAN setup, the security function is configured by the SD-WAN 
controller to have one or more virtual links for particular target networks.  
As an example, the NSF might have three ways of getting to subnet 192.0.2.0/24
:
Using an MPLS link. This is the most expensive way
Using the Internet through an LTE modem. This is moderately expensive as well.
Using the Internet through a DSL link. This is the cheapest way. If at all 
possible, you would want to use this.

An SD-WAN gateway uses some measurement technique to evaluate some properties 
of each one of these physical links: RTT, jitter, packet loss.

The SD-WAN gateway receives policy from the controller. Such a policy may 
require that VoIP traffic go over the cheapest link where RTT is no more than 
50ms and jitter is no more than 50ms. Video traffic might be more stringent on 
jitter but less stringent on RTT. Web transactions are fine with any RTT and 
jitter as long as packet loss is small. There’s more to it and you can find 
more with a Google search and at the vendors’ websites. One part is dynamic 
discovery of peers for large environments.

Most of SDN started in academia and migrated to the “real world”. In this 
particular case it seems that private industry is somewhat ahead. I don’t think 
we’d want to standardize a partial solution here. It’s probably better for us 
to try to get all of the functions that people expect under the buzzword 
“SD-WAN”.

I’m not sure either whether this is work for this working group or whether it 
justifies its own working group. I’m also not sure how we can attract the 
SD-WAN vendors to the IETF to participate. But I think we should be “thinking 
big” on this.

Yoav

_______________________________________________
I2nsf mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/i2nsf

Reply via email to