Making it to OSPF WG, finally.
------------------------------
*From:* Gregory Mirsky
*Sent:* Wednesday, March 20, 2013 10:16 AM
*To:* 'Alia Atlas'
*Cc:* Clarence Filsfils; [email protected];
[email protected]; [email protected]; [email protected];
John E Drake; [email protected]; [email protected]
*Subject:* RE: [Isis-wg] IGP-TE extensions for unidirectional Link Delay,
Delay Variation, and Link Loss
Hi Alia, et al.,
I was thinking about reflecting queuing without actually measuring it. I
think that major concern with measuring dynamic queuing is that we have
only active, including synthetic, measuring methods that change the
process, at least that is to the best of my knowledge. But there're
discussions on passive and/or hybrid methods that will be less distractive
to the traffic. But there's another approach that I see as most practical
at this time - AQM. AQM can be expressed in maximum queuing delay and it
will exactly characterize unidirectional delay, delay variation on the
given link.
Again, it might not happen tomorrow but I believe that it is beneficial to
make tool flexible for the future, perhaps near future.
Regards,
Greg
------------------------------
*From:* Alia Atlas [mailto:[email protected]]
*Sent:* Wednesday, March 20, 2013 2:05 AM
*To:* Gregory Mirsky
*Cc:* Clarence Filsfils; [email protected];
[email protected]; [email protected]; [email protected];
John E Drake; [email protected]; [email protected]
*Subject:* Re: [Isis-wg] IGP-TE extensions for unidirectional Link Delay,
Delay Variation, and Link Loss
I agree with Spencer that I don't see the extra information on a per-queue
being useful if we are not measuring dynamic queuing behavior and delay.
How do you see these values being different? Where would you get them
from while avoiding the concerns about a traffic-dependent feedback loop?
Alia
On Mar 19, 2013 3:14 AM, "Gregory Mirsky" <[email protected]>
wrote:
> Dear Authors, et al.,
> As mentioned at the mike during IS-IS WG meeting in Orlando, I believe
> that three parameters, Link Delay, Delay Variation, and Link Loss are per
> priority/Traffic Class. Other link characteristics introduced in
> draft-previdi-isis-te-metric-extensions- and
> draft-ietf-ospf-te-metric-extensions- documents, Residual Bandwidth and
> Available Bandwidth, might not have distinct per priority connotation.
>
> I think that per priority advertisement does not imply dynamic queuing
> being included, not suggest how these metrics get their values (dynamic
> measurement or estimate), but allows more realistically characterize a link.
>
> Comments, suggestions are welcome and greatly appreciated.
>
> Regards,
> Greg
>
>
> _______________________________________________
> Isis-wg mailing list
> [email protected]
> https://www.ietf.org/mailman/listinfo/isis-wg
>
>
_______________________________________________
OSPF mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/ospf
