Re: [dhcwg] RFC 3315 message re-transmission query

[Johan Bosaeus <johanb at weird-solutions.com>]

On Tuesday 03 November 2009, ravi kumar wrote:
> Thanks a lot to people in forum for promptly answering my queries.
> I have one more query regarding calculation of Retransmission time-out in
> Section-14 of RFC 3315:
>
>  RT for the first message transmission is based on IRT:
>       RT = IRT + RAND*IRT
>    RT for each subsequent message transmission is based on the previous
>    value of RT:
>       RT = 2*RTprev + RAND*RTprev
>
> Do RTprev value used in subsequent calculation of RT, include randomization
> added by RAND ?
> I guess should not be the case, because the retransmission algorithm is
> exponential back-off algorithm and incase RAND is added multiple levels in
> calculation of RT,this exponential nature is lost.

I don't think the exponential back-off is really lost. It's just that the 
exponential back-off is overloaded with a randomization that is sort of 
proportional to the backoff. I.e. the more attempts you do, the larger the 
back-off and the larger the randomization is.

> I have done simple Math in calculating maximum and min waittimes for
> REQUEST message, by considering "RTPrev includes RAND" :
> 0.9 ---   1.1
> 1.711 --- 2.31
> 3.249 --- 4.851
> 6.1731 -- 10.1871
> 11.72889 --- 21.39291
> 22.284891 --- 44.925111                            -> The time difference
> here is very large here

Isn't is so that the standard recommends a maximum of 30 seconds backoff. From 
that point the maximum spread is +-3 seconds.

- Johan

-- 
Johan Bosaeus
Weird Solutions
tel: +46 8 758 3700
fax: +46 8 758 3687
http://www.weird-solutions.com