Hi Eric,
Pls see inline with [weiguo].
Thanks
weiguo
________________________________________
From: Erik Nordmark [[email protected]]
Sent: Tuesday, November 18, 2014 2:40
To: Haoweiguo; Larry Kreeger (kreeger); Greg Mirsky
Cc: [email protected]
Subject: Re: [nvo3] 答复: 答复: Comments on NVO3 data plane requirements for OAM
On 11/13/14 11:31 PM, Haoweiguo wrote:
> I would prefer the marking bit in NVO3 header, rather than in outer IP header.
> This is for overlay network performance measurements, not for underlay
> network.
My understanding is that you want to measure from the encapsulating NVE
to the decapsulating NVE. Is that correct?
[weiguo]: Right.
The outer IP+NVO3 headers travel from NVE to NVE, thus I don't
understand why we should consider the fields in that outer header
differently than the fields in the NVO3 header.
> In ingress NVE, marking policy should support discrimination between
> different tenants,even support discrimation different applications of same
> tenant.
Are you concerned about the DSCP bit? (Those are different than the ECN
bits).
[weiguo]: Sorry, after i discussed with some experts, the 2 bits can be in
outer IP header or NVO3 header, both methods can discriminate the flow of
different tenants. But outer IP header has no extra space for these two bits,
ECN bits had better not be multiplexed. So the two marking bits had better be
included in NVO3 header.
Erik
> In this case, marking bit can only be set in NVO3 header. And just as Mach
> and Greg's description, two bits are necessary, one bit for packet loss
> detection, another bit for packet latency detection.
> Thanks
> weiguo
_______________________________________________
nvo3 mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/nvo3
