[e2e] An alternative algorithm of calculating RTO
jaglee at peoplemail.com.cn
Mon Jun 3 18:28:27 PDT 2002
>>Maximum value of N can be aribitually chosen. It should not be less than 8 nor
> > greater than 32.
>Isn't the limit of 8 a problem -- most transactions don't last that many
>round-trips. And how much more accurate is this estimate (in particular,
>is the probability it is too short -- the most important issue -- greater
>or less than that of the old calculation)?
Well, there're two hidden assumption:
(1) The mrtts of the most recently acknowledged N packets without re-sending approximately reflect the current state of the network can we say rartt and rvrtt.
(2) For transactional communication, congestion control may be of little use. I mean the estimate is intent for long persistent connection.
While assumption 2 may be acceptable, assumption 1 is really a problem.
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 2015 bytes
Desc: not available
Url : http://www.postel.org/pipermail/end2end-interest/attachments/20020604/a2b63100/iWPS4362.bin
More information about the end2end-interest