Re: [RPRWG] Error in rate normalization?
Fredrik,
localFairRate is defined as number of bytes per agingInterval. AgeCoef is used to smooth it. It is not in the unit
of it.
Thanks.
Necdet
Fredrik Davik wrote:
> I am having troble interpreting the transformation of localFairRates
> to a rate that can be interpreted by other stations on the ringlet
> (normalized rate) and the way this is done in the draft
>
> Either I am having a bad day in office or else, something is wrong ;)
>
> Take a look below:
>
> localFairRate is specified in units of:
>
> 1. localWeight * [bytes/(ageCoef * agingInterval)]
>
> 2. normCoef = localWeight * rateCoef * ageCoef
>
> Where ageCoef and localWeight are per station/ringlet configured
> parameters and agingInterval is a global setting which depend on the
> link rate.
>
> The normalization process aims to distribute the rate information in a
> way that is not dependent on locally configured parameters, thus
> dividing 1.(localFairRate) above by 2. (normCoef) :
>
> We get :
>
> 3: ( localWeight * [bytes/(ageCoef * agingInterval)]) / (localWeight * rateCoef * ageCoef)
>
> 3. result when reorganized and simplified:
> bytes / ( ageCoef * ageCoef * agingInterval * rateCoef)
>
> >From this I observe two things:
>
> a) The "normalized" value still contains the effect of the local
> ageCoef parameter
>
> b) agingInterval is the same for all stations and could also be
> removed from the expression by including it in normCoef:
>
> If normCoef is rewritten as:
>
> 4. normCoef = (localWeight * rateCoef)/ ( ageCoef * agingInterval)
>
> What we get when diving 1 by 4 is:
>
> 5. ( localWeight * [bytes/(ageCoef * agingInterval)]) / ((localWeight * rateCoef)/ ( ageCoef * agingInterval))
>
> 5. result when reorganized and simplified (this is the rate
> distributed on the ring):
>
> bytes/rateCoef
>
> When the normalized rate is received by another station, this is
> converted to a local rate by multiplying the normalized rate by
> normCoef (the modified one above) as before.
>
> Best Regards,
> Fredrik
> ---------------------------------------------
> Fredrik Davik
> Phone: +47 67 82 83 88
> Mobil: +47 45 24 91 88
> Fax: +47 67 82 82 01
> Switchboard: +47 67 82 82 00
>
> mailto:bjornfd@xxxxxxxxx
>
> http://heim.ifi.uio.no/~bjornfd
> http://www.simula.no/people_one.php?people_id=22
>
> Simula Research Laboratory
> P.O.Box 134, Lysaker
> N-1325 Lysaker
> Norway