Jump to content

Talk:Weighted fair queueing

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia
(Redirected from Talk:Weighted fair queuing)

Generalized processor sharing

[edit]

I do not understand why User:Kvng has removed the add I have done (add of the name "PGPS" and WFQ algorithm), making reference to Generalized_processor_sharing (GPS).

The WFQ is also called Packetized GPS, as explained in [1]. WFQ is currently presented has a "generalization of fair queuing", but it also is a generalisation of Generalized processor sharing. And currently, the description of WFQ just presents it has a generalisation of FQ allowing to give an average rate of . But this is also the aim of GPS, and WRR and Deficit_round_robin (DRR).

In fact, WFQ, WRR, DRR are all packet based implementations of the GPS idea of rate-based sharing of capacity.MarcBoyerONERA (talk) 08:51, 16 January 2014 (UTC)[reply]

If we're going to include aliases in the lead, there usually should be a WP:REDIRECT associated with them. If we created a redirect for Packetised Generalized Processor Sharing it would probably better direct readers to Generalized processor sharing.
There is clearly room for improvement in the organization of the topics you're mentioning here. I reverted the edits because I did not see how these were incremental improvements; in my assessment, they defocused this article. If you have some larger reorganization in mind, of which this was just a first step, let's hear about it here. ~KvnG 17:43, 19 January 2014 (UTC)[reply]
Yes, I think that the pages on WFQ, WRR, DRR and GPS must be re-organised. GPS is a general idea, trying to share the bandwidth/throughput between flows, with user-defined ratios. The algorithms WFQ, WRR and DRR are three implementations of this principle. Such links must be clear. But how to do it? One unique page on GPS and its implementations? On page on GPS, and one per implementation (as the current state), augmented with links between all? MarcBoyerONERA (talk) 14:23, 29 January 2014 (UTC)[reply]
I think we're now in good shape at least in the lead WRT this issue in this article. Things are not so clear over at Fair queuing but I think we can close this discussion here and work on that over there. ~Kvng (talk) 00:38, 6 March 2018 (UTC)[reply]

References

  1. ^ Parekh, A. K.; Gallager, R. G. (1993). "A generalized processor sharing approach to flow control in integrated services networks: The single-node case" (PDF). IEEE/ACM Transactions on Networking. 1 (3): 344. doi:10.1109/90.234856.
[edit]

The following comment was posted to my talk page. I believe this refers to this edit of mine. I think it best to discuss here instead.

you have simplified the reference to the "Fair queuing principle" to a simple mention to "fair-queuing scheduler". My point of view is that the term "fair queuing" has two meaning: a general principle and an specific scheduling policy. Then, the expression "all fair-queuing schedulers" is ambiguous. This is why the term "Fair queuing principle" was used, with a reference to the section. MarcBoyerONERA (talk) 05:36, 15 April 2015 (UTC)[reply]

This is something that is best conveyed in the copy and not as a (subtle) section link. Before proceeding with improvements here, please read my recent request at Talk:Fair_queuing#Family_of_Algorithms ~Kvng (talk) 19:28, 15 April 2015 (UTC)[reply]