Talk:Queueing theory

From Wikipedia, the free encyclopedia
Jump to: navigation, search

Limitations of the Mathematical Approach[edit]

This section needs references to avoid the original research claim imo. It almost is a point of view of an author injected into the document. Consider the supporting argument: "For example; the mathematical models often assume infinite numbers of customers, infinite queue capacity, or no bounds on inter-arrival or service times, when it is quite apparent that these bounds must exist in reality." That doesn't really support the point of view "limitation of mathematical approach" because these things can be tackled within a mathematical framework, there is nothing intrinsic to mathematical modeling that forms a limitation that keeps these ideas out. It is likely: they haven't been distilled down to a end user level where they exist in mathematical literature, they haven't been done yet, or they have intractable pieces of them waiting to be solved. Either way that section needs a rewrite to be based more on peer reviewed literature and less on the authors opinion. —Preceding unsigned comment added by (talk) 17:18, 26 March 2010 (UTC)

Agreed, this isn't true of all mathematical modeling of queueing. There is much mathematical framework built to handle all sorts of these things claimed not to be feasible mathematically. -unsigned —Preceding unsigned comment added by (talk) 06:11, 10 April 2010 (UTC)

This happened a long time ago, but I think that it's a great pity that the article lacks a 'criticism' section. There are several issues with queueing theory when applied to the real world. It certainly has its value, when applied to appropriate situations, but it also has very severe limitations. The article is now misleading in suggesting that it is an established fact that queueing theory not only works, but is practically useful, with no caveats at all.

I understand that you don't want original research in wikipaedia. That's quite reasonable. However, it is a fact that queuing theory is hopeless if you wish to get an estimate of the user experience of response time in a complex system - only a simulation actually works.

I fear that the article may mislead people into making dangerously false assumptions.

Fustbariclation (talk) 10:51, 10 July 2014 (UTC)

I had a few minutes to have a look on google scholar. Here's an article that points out that

I've marked the article as unbalanced - I think that a criticism section that points out the limits of applicability of the theory would improve the article considerably. Fustbariclation (talk) 11:05, 10 July 2014 (UTC)

I agree a section on limitations may be useful, do you have any reputable sources discussing this in mind? Gareth Jones (talk) 13:34, 10 July 2014 (UTC)
I rather hoped that somebody else might. I have some sources in mind, but I'm not sure if they're reputable enough for here. I'll have a look around to get some good ones. Fustbariclation (talk) 06:02, 11 July 2014 (UTC)

Here's an article that points out that queuing theory isn't that good for semiconductor [Queueing theory for semiconductor manufacturing systems : A survey and open problems]. It also proposes some solutions. It took a few minutes in google scholar to find this. I'm sure that there is lots more if anybody is interested in having a look.

Also Salt, John D. "Simulation should be easy and fun!." Proceedings of the 25th conference on Winter simulation. ACM, 1993. talks of needing simulation where "live system is impractical; where linear programming is insufficient; where queueing theory is inadequate'.

This also is relevant [[ Modelling patient flows and resource provision in health systems R Davies]] the abstract makes the limitations, relative to simulation, clear "Patient flow models may be used for planning health services for both acute and chronic patients. There are models which assume sub-groups of patients are homogenous and events occur at equally spaced intervals of time. These include Markov and semi-Markov chain models, queueing models and deterministic models of the transition of patients between states. These techniques are useful for examining patient flow in large population groups where Markov assumptions, or simple extensions of these, can be made. Discrete event simulation models allow patients to have individual attributes and to interact with resource provision but they are more time consuming to test and run. They are particularly suitable for models of systems of patient care where the constraints on resource availability are important. They may also be used on unconstrained population models with several thousands of patients. A significant development in simulation is the facility to model entities so that they can participate in more than one activity simultaneously and interrupt each other. The credibility of any model is dependent on reliable data which are not always readily available in the British Health Service"

Fustbariclation (talk) 11:05, 21 July 2014 (UTC)

Buzen's algorithm[edit]

There's a fragment of an article on Buzen's algorithm that I'm working on with a few others, but when we tried to move it here we moved it to Talk:Queueing_Theory instead by accident. So, this comment is a placeholder until it finds its way over here. John Reed Riley 04:29, 19 March 2006 (UTC)

I've moved the fragment over to Buzen's algorithm, despite the fact that it isn't finished yet. Hopefully someone else has enough time to finish it. John Reed Riley 19:44, 11 November 2006 (UTC)

Spelling: "queueing" versus "queuing"[edit]

I've reverted a recent edit that changed the spelling of "queueing" in the article to "queuing". My reasons are as follows. Firstly, all literature on queueing theory that I have seen uses the "queueing" spelling. Also, the name of the article is "Queueing_theory". There's even a link in the article to a FAQ explaining that "[the] vast majority of queueing theory researchers use 'queueing.'" John Reed Riley 21:42, 18 April 2006 (UTC)

The wiktionary entry for "queuing" gives "queueing" as an alternative spelling. The cited page, for queueing, doesn't list an alternative spelling. That could well be construed as evidence that the spelling "queueing" is the primary form, as this page implies. That's a bit tenuous, though, especially when only one of the two wiktionary entries is cited. Shouldn't the citation refer to the FAQ entry mentioned above? Mind, it timed-out on me just now. is another possible source. Ah, I see the page cited above has moved under "/math" - Now I've found it, it does look persuasive. Being Google's top match for both these words is likely to lead people here when they're more interested in the spelling than the theory.

Since when is popular vote the criterion for "correct spelling"? Shouldn't an article of an encyclopaedia, if indeed wishing to focus on this topic, simply cite a dictionary and/or a grammar book? As an example, here's the entry at Merriam-Webster

He's right, though, that virtually every paper and textbook on the subject uses "queueing", so that must be considering the industry standard. [1]

Actually the cited Merriam-Webster link provides both spellings.

There should at least be a citation for a claim like this, since the dictionary says that both spellings are valid (I've added a 'citation needed' to it.), however there is a Gordian solution here: The entire etymology section is not relevant to the article, and really adds no value. It should probably be removed. Thoughts?

I've added back the etymology section but titled it spelling and removed the bit on the etymology of queue. It could be reworded but there is clearly something needed here to say why the section is titled queueing not queuing.Olsenprof (talk) 05:07, 5 June 2016 (UTC)

Rude Customer Handling[edit]

It is incorrect to say there is no math modeling of "rude" customers such as network traffic flows. I did some back in the mid 90's and the work was continued by my former research partner.

Hmmm.. does the article imply this anywhere? I couldn't see where it did. Does "Rude" here mean anything other than "preemptive" (pushing in)? That's a standard part of queuing theory hand has been for some while. --Richard Clegg 07:26, 17 May 2006 (UTC)

In the limits of the math approach section it is implied... by rude I mean they arrive out of order and/or not at all (there are missing arrivals and what does arrive is not in time order)

The "limits of the math" section really says nothing about that issue whatsoever. --Richard Clegg 21:52, 17 May 2006 (UTC)
What does one really mean by rude customer handling? What is polite customer handling, for that matter? I think using terms such as rude or polite are too emotive and vague for a scientific theory. These terms should be avoided as they describe some person's opinions about the way customers are handled, not the specific way customers are handled. For example: First Come First Served describes the way arriving customers are handled. Some may consider this polite, in some situations. However, if you were seeking service for a life threatening situation, being delayed by someone who did not have a more serious situation would be rude - to you - here service should be Most Serious Served First. Thus rude is a subjective opinion, while describing the method of service is a more objective assessment. And if everyone understands and accepts the service regime then it is not rude, either - it is the unexpected that becomes rude. -- Cameron Dewe 00:20, 1 October 2006 (UTC)
Only this section of the talk page mentions "rude" customers. I can only assume it is a reference to pre-emption. I've never heard it used in a technical sense. --Richard Clegg 12:18, 1 October 2006 (UTC)

And the point is?[edit]

This article fails to actually explain queueing theory!

Yes -- that would take a book. Probably a book which only someone with at least an undergraduate training in a mathematical area could understand. However, if there are specific omissions which could be addressed then I'd like to work on this. --Richard Clegg 15:11, 28 September 2006 (UTC)
Many books and articled have been written on the subject. I think the issue is the article barely scratches the surface of the subject as it fails to explain that queues are encountered in so many different situations it is difficult to even recognise it is a queue, let alone understand that queueing theory applies to the situation. My particular criticism of the article is that it is too detailed and piecemeal about specific situations without first laying down some basic groundwork about how common queueing is in our lives and hence how widely queueing theory is applicable. -- Cameron Dewe 00:20, 1 October 2006 (UTC)
I like what you've done to the article. It's definitely improved. --Richard Clegg 12:18, 1 October 2006 (UTC)
"In queueing theory a model is constructed so that queue lengths and waiting times can be predicted." I think this is good enough to give the read an idea about why people bothered studying queueing theory. Cristiklein (talk) 11:02, 28 January 2013 (UTC)

Link Spam[edit]

I posted a link to JMT some time ago but it was removed as "link spam"... I don't think it is "spam", as it's an open source software, GPL, that is used to solve queueing network models either using analytic, or asymptotic or simulation techniques. It's exactly the kind of instrument useful to learn queueing network theory... If you think that link should not be placed in that page remove it but please give a right motivation and not tag it as "spam"... —Preceding unsigned comment added by Bertoli (talkcontribs) 19:15, 26 April 2007 (UTC)

History section[edit]

Can we expand this? I think some names like Erlang, John Little, Jackson, Kendall, Buzen and Markov are worthy metioning. Does anyone know some bibliographical source? — Preceding unsigned comment added by Lbertolotti (talkcontribs) 17:25, 6 August 2012 (UTC)

k now most important stuff is covered and sourced, maybe the refimprove at the beginning may be removed now?Lbertolotti (talk) 17:37, 20 April 2013 (UTC)


Weird notation[edit]

In Section Queueing_theory#Utilization, M is used to denote the number of servers. Shouldn't one use k instead, as presented in the Queueing_theory#Single_queueing_nodes Section? — Preceding unsigned comment added by Cristiklein (talkcontribs) 10:59, 28 January 2013 (UTC)

I've corrected this (using c) to match the M/M/c queue article. Gareth Jones (talk) 12:37, 28 January 2013 (UTC)

Scope of applicability of queueing theory to real problems[edit]

Someone added this tag to the page: it is a fundamental issue, but does anyone have sources on this matter?Lbertolotti (talk) 05:53, 27 March 2015 (UTC)

M/G/k and G/G/k[edit]

So nobody has ever managed to analytically solve the M/G/k and G/G/k models?Lbertolotti (talk) 02:08, 30 March 2015 (UTC)

Correct, the general problem is hard! See the M/G/k queue article, which includes the quote from Tijms et al. that it is "not likely that computationally tractable methods can be developed to compute the exact numerical values of the steady-state probability in the M/G/k queue." The G/G/k queue is a further generalisation. Gareth Jones (talk) 21:43, 8 September 2015 (UTC)

Example of M/M/1 in Queueing Networks section[edit]

Imho, the example in the queueing networks section is neither fitting to the section (since it is a derivation for a single queue) nor it is written in a comprehensible way. I propose to move and revise this example. --Lippofant (talk) 12:04, 20 April 2017 (UTC)