Re: [tsvwg] Some comments on NQB (part 2)

[[e-mail]]

Jason,

You write: "But IMO to achieve the greatest level minimize the need to more tightly couple different networks' respective policies & operation."

From my experience of integrating and modifying QoS work from interconnection/service area across a long term DiffServ aware backbone and access network to home-gateway for a couple of years, I'm sure, unless you tightly couple different network sections respective QoS policies & operation, you are bound to fail. 

Regarding where and why to re-mark or not to re-mark DSCPs, please read RFC3260, I pasted the text into
https://mailarchive.ietf.org/arch/msg/tsvwg/_q585aV_UgEZbm5QqtEHbotE3UM/

Regards, Ruediger



-----Ursprüngliche Nachricht-----
Von: Livingood, Jason <[e-mail]> 
Gesendet: Mittwoch, 4. Mai 2022 15:14
An: Greg White <[e-mail]>; Gorry Fairhurst <[e-mail]>; Geib, Rüdiger <[e-mail]>
Cc: [e-mail]
Betreff: Re: [tsvwg] Some comments on NQB (part 2)

>        - However, I also would like clarity for the case where there is no
        specific negoiation SLA, then I think the traffic SHOULD use DSCP 45,
        and ought to then pass this codepoint unmodified through this diffserv
        domain. My rationale is that we ought to avoid adding a remarking
        behaviour for routers that could transparently forward the traffic.

[JL] +1 to "we ought to avoid adding a remarking behaviour for routers that could transparently forward the traffic." It seems best (most efficient, least administrative work/cost, fastest) to be able to use DSCP 45 across network boundaries as a default behaviour. If some other network does not want to permit marking or wants to use DSCP 5, then I guess we can talk about it ("negotiate"); hopefully these cases are small in number. But IMO to achieve the greatest level minimize the need to more tightly couple different networks' respective policies & operation.

    [GW]  For a transit network, the value 5 could be used to transparently forward traffic without re-marking.

[JL] Why would a transit network use 5 instead of 45? In any case, I would not use transit networks as the primary use case - because by traffic volume the majority of traffic to access networks likely flows over direct interconnect links rather that via transit routes.