Björn Grönvall <[email protected]> writes:

> This is definitely confusing and even Varghese (one of the DRR authors) uses 
> this analogy:
>
>     A banking analogy motivates the solution. Each flow is given a quantum,
>     which is like a periodic salary that gets credited to the flow’s bank
>     account on every round-robin cycle. As with most bank accounts, a flow
>     cannot spend (i.e., send packets of the corresponding size) more than is
>     contained in its account; the algorithm does not allow bank accounts to
>     be overdrawn. However, perfectly naturally, the balance remains in the
>     account for possible spending in the next period. Thus any possible
>     unfairness in a round is compensated for in subsequent rounds, leading
>     to long-term fairness.
>
> Quote is from the book Network Algorithmics (p. 351) by George Varghese.
>
> Perhaps the banking analogy would be less confusing?

Yeah, that's why I switched to using 'credit' as the term for what is
being kept track of in the next revision:
https://kau.toke.dk/ietf/draft-ietf-aqm-fq-codel-04.html

Can't really change the fact that the algorithm is called 'deficit
round-robin', though, so can't get rid of the 'deficit' term entirely.

-Toke

_______________________________________________
aqm mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/aqm

Reply via email to