Talk:Proportional-fair scheduling

Accessable?
I tried to make this document more readable to a general audience. Let me know if i succeeded. Thanks! --Mblumber (talk) 03:03, 26 January 2008 (UTC)


 * Thanks for the improvement!
 * The problem with this article is that the formulas and terminology used in section 1 and 2 differ. How are the WFQ weights and the priority function related to each other? I would also be glad if the same terminology was used to explain the difference between proportional fair, max-min fairness, maximum throughput scheduling and Channel-dependent scheduling. There should be an article on scheduling (communication).  Mange01 (talk) 21:52, 3 March 2008 (UTC)

Exponents
In all references I know about proportional fair, the expression $$P=\frac{T^\alpha}{R^\beta}$$ appears with $$\alpha=\beta=1$$. The generalization to $$\alpha, \beta$$ arbitrary is interesting. However, I have not found it in any of the two referenced papers. In which paper is this version of proportional fair (i.e. that with arbitrary exponents $$\alpha,\beta$$) introduced or discussed?

Another question is: should "proportional fair" denote the general expression, or the one with $$\alpha=\beta=1$$)? As I said, all references I read call the version without exponents "proportional fair". Perhaps the paper that introduced the exponents gave that version some name; or else could it be called "generalized proportional fair"?

Lmendo (talk) 10:50, 17 September 2010 (UTC)


 * I looked through all the sources that I have on the subject, and I cannot find and reference to variable $$\alpha, \beta$$ being other than unity. Looking at the very oldest version of this page, that fact was included. I cited it with a paper that I found, but when I review that paper I cannot figure out why I cited it. I'll keep digging however, because I'm curious now. --Mblumber (talk) 01:35, 20 September 2010 (UTC)

Weights
The article states that the weights are w_i = 1/c_i, where c_i is the cost of consumed resources per data bit. However this is not true of the "Proportional Fair Scheduling" algorithm used in CDMA HDR (high data rate) systems (next section). However according to David Tse https://web.stanford.edu/~dntse/papers/ima810.pdf: (quote) "Round-Robin Policy • Give same number of time slots to all the users in a round-robin fashion, regardless of their channel conditions. Proportional fair policy: • Give roughly the same number of time slots to all users, but try to transmit to a user when its channel condition is near its peak" (end quote). In the above same = round robin. So w_i = 1/N, where N is number of users. If w_i is proportional to SNR_i, then the throughput achieved will be ratio of SNR_i^2. However, I have seen this w_i = 1/c_i in a few other places and hence am not sure how to edit this article. — Preceding unsigned comment added by IrfanAli04 (talk • contribs) 13:18, 10 October 2014 (UTC)