Re: [DMM] Presentation of https://datatracker.ietf.org/doc/draft-zzhang-bess-ipvpn-payload-only/

Jeff Tantsura Tue, 08 Nov 2022 09:13:16 -0800

Jeffrey,

UDP port range is used as an identifier (not as mutable field/ input into hashing).

Not all mobile networks use MPLS, even less MPLS networks actually use entropy labels (all routing silicon (within its capabilities)is trying to look beyond MPLS label stack anyway).

Cheers,

Jeff

On Nov 8, 2022, at 16:49, Jeffrey (Zhaohui) Zhang <[email protected]> wrote:

Hi Uma,

Indeed, the two drafts use GTP-U as an example use case but the mechanism can be used for general use cases and it is already stated as such.

If transit routers use UDP ports for ECMP hashing, I assume they’d expect full IP header before the UDP header. Therefore, the entire idea is out. Supposedly, MPLS has its own entropy label mechanism for transit routers to do hashing (vs. relying on deep packet inspection of 5-tuple), and that would be used in this case.

I will look into tn-aware-mobility draft.

Thanks.

Jeffrey

Juniper Business Use Only

From: Uma Chunduri <[email protected]>
Sent: Saturday, November 5, 2022 6:39 AM
To: Jeffrey (Zhaohui) Zhang <[email protected]>
Cc: [email protected]
Subject: Re: [DMM] Presentation of https://datatracker.ietf.org/doc/draft-zzhang-bess-ipvpn-payload-only/

[External Email. Be cautious of content]
"For example, PE1 can advertise a label for a (source, destination,
   IP/UDP payload type) tuple with the local semantics being that
   incoming traffic will be encapsulated in an IP or IP+UDP header and
   then routed out.  When PE2 receives IP or IP+UDP traffic from the
   UPF, if there is a label for the corresponding (source, destination,
   IP/UDP payload type) tuple, it removes the IP or IP/UDP headers and
   simply transport the remaining payload.  In this 5G scenario, it is
   still GTP - just that the IP/UDP headers are not present between PE1
   and PE2.
"

Very useful.

And this doesn't have to be tied to GTP overlays. I would recommend generalizing this (just keep the overlay header intact) - though you can point to GTP as an example.

UDP might be needed for load balancing the traffic in the transport network. So better to keep this intact.

UDP Src port (is one way) to encode the slicing information as specified in https://datatracker.ietf.org/doc/html/draft-ietf-dmm-tn-aware-mobility

Thx!

--

Uma C.
On Thu, Jul 28, 2022 at 6:08 PM Jeffrey (Zhaohui) Zhang <zzhang=[email protected]> wrote:

Hi,

Due to a glitch the slides for the above draft were not available so it was not presented as planned in the DMM session in IETF114.
However, it was presented in the BESS session and the following are the video recording and slides:

https://youtu.be/V2r68JhrQag?t=660
https://datatracker.ietf.org/meeting/114/materials/slides-114-bess-draft-zzhang-bess-ipvpn-payload-only-00

The direct use case that triggered the draft is GTP-U transportation, so I hope it is of interest to this group. Appreciate your comments.

Thanks.
Jeffrey

Juniper Business Use Only

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

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

Re: [DMM] Presentation of https://datatracker.ietf.org/doc/draft-zzhang-bess-ipvpn-payload-only/

Reply via email to