[RPRWG] maint. request 47 - discussion ?
Hi John,
Are any of those offline discussions documented ? I wanted to read through
them, if they are accessible.
I think the remedy suggested may have a small anomaly - there can be a
condition where (lpNrXmitRate > rateLowThreshold) is TRUE but
(lpNrXmitRate -
lpAddRate > rateLowThreshold) is FALSE, even when lpAddRate is small. In
such a case, we would be facing the same problem. How about having
"(lpNrXmitRate - lpAddRate > factor*rateLowThreshold)" as the condition
instead, where 'factor' with range (0,1), is such that the latter condition
is FALSE only when lpAddRate is sufficiently large enough.
Also, I was wondering if there is actually a need to remedy this problem.
This is only a utilization issue and can accept this as a disadvantage of
using aggressive method. However, if we do want to remedy this - why not
also look towards active weights to give a better estimate of localFairRate
?
Mohit
> ----- Original Message -----
> From: "J Lemon" <jlemon@ieee.org>
> To: <mohitsood@ieee.org>
> Cc: <STDS-802-17@listserv.ieee.org>
> Sent: Tuesday, August 02, 2005 12:24 PM
> Subject: RE: [RPRWG] maint. request 47 - discussion ?
>
>
> Mohit,
>
> The proposed remedy was provided in a separate file (since it didn't fit
> in the space provided in the form). There is a link to the additional file
> from the Detailed History view. [Note that several MRs have additional
> files associated with them, all linked from the Detailed History view.]
>
> The only discussion so far was that people wanted to consult offline
> before making a decision, hence the status of T.
>
> Your suggestion below appears to address conservative mode, whereas this
> is for aggressive mode.
>
> jl
>
> -----Original Message-----
> From: Mohit Sood [mailto:msood@UNITY.NCSU.EDU]
> Sent: Tuesday, August 02, 2005 7:18 AM
> To: STDS-802-17@listserv.ieee.org
> Subject: [RPRWG] maint. request 47 - discussion ?
>
>
> Hi,
>
> Maintenance request no. 47, states - "In the case of the aggressive
> fairness method, a station may become the head of a congestion domain
> based solely on the rate of transiting non-reserved traffic, even when the
> local station is not adding fairness-eligible traffic or when transiting
> fairness-eligible traffic is sufficient to exceed the rateLowThreshold
> without the contribution of the local station. In such cases, add rates
> within the congestion domain are unnecessarily (and repeatedly) throttled
> to a small value, significantly reducing ring utilization."
>
> No remedy has been provided by the person who has submitted this
> maintenance. I wondering if there has been a discussion about the problem
> mentioned. How about if /localFairRate /is set to /(lpAddRate +
> lpFwRate)/2 /when /activeWeightsDetection /is disabled and /(lpAddRate +
> lpFwRate) * (localWeight /activeWeights)/ when /activeWeightsDetection /is
> enabled?
>
> Mohit Sood
> Grad Student (Dr. Viniotis)
> NC State University
>