Hi Robert
Thanks, I agree with you that one needs some form of „historical load state” to
detect “massive congestion” on a backbone interface.
---
RR: If I have massive congestion I may really want to "protect" priority
flows and only trigger it when priority class get's full.
RG: How do you determine “massive congestion” if the majority of the traffic is
rate controled? If your above statement only refers to a queue or sub-class
usually not experiencing loss, that would clarify things to me.
RR: Right I did mean just take 10 [s] or 60 [s] or time T [s] queue depth
averages and compare with historical data. If it is N x higher then perhaps we
need to take some action. And by some action I am not sure if TTE would be best.
---
RG: In our backbone, we engineer traffic by metric optmisation (and have
suitable machinery in place). I think or hope at least that Segment Routing
provides a toolset to swiftly route a portion of traffic around a congested
bottleneck (supported by the “machinery in place”).
Regards,
Ruediger
Von: Robert Raszuk <[email protected]>
Gesendet: Montag, 3. April 2023 10:40
An: Geib, Rüdiger <[email protected]>
Cc: [email protected]
Betreff: Re: TTE
Hi,
while I’m not involved in this discussion, I’d still have one question to you
Now you are :)
You write: If I have massive congestion I may really want to "protect"
priority flows and only trigger it when priority class get's full.
How do you determine “massive congestion” if the majority of the traffic is
rate controled? If your above statement only refers to a queue or sub-class
usually not experiencing loss, that would clarify things to me.
I think we are talking about a network where the traffic rate is not ingress
controlled. If it is, I don't think it is likely to get congestion which TTE
could mitigate for unicast flows.
Right I did mean just take 10 [s] or 60 [s] or time T [s] queue depth averages
and compare with historical data. If it is N x higher then perhaps we need to
take some action. And by some action I am not sure if TTE would be best.
Personally I would rather see fix being applied on the end to end path not on
link by link in the IGP core.
Regards,
R.
_______________________________________________
rtgwg mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/rtgwg
