[go: up one dir, main page]
Wikipedia

G-network: Difference between revisions

Browse history interactively
← Previous edit
Content deleted Content added
VisualWikitext
Egelenbe (talk | contribs)
70 edits
No edit summary
m Reverted 1 edit by Lamayomer (talk) to last revision by Onel5969
 
(48 intermediate revisions by 25 users not shown)
Line 1:
{{otheruses4about|the mathematical model|the wireless standard|802.11g}}
In [[queueing theory]], a discipline within the mathematical [[probability theory|theory of probability]], a '''G-network''' ('''generalized queueing network''',<ref>{{cite journal | doi=10.2307/3214499 | title = Product-form queueing networks with negative and positive customers | first = Erol | last = Gelenbe | journal=Journal of Applied Probability | volume = 28 | number = 3 | date = 1991 | pages = 656–663 | url=https://www.di.ens.fr/~busic/mar/projets/G91.pdf}}</ref><ref>{{cite journal | doi = 10.2307/3214781 | title = G-Networks with Triggered Customer Movement | first = Erol | last = Gelenbe | authorlinkauthor-link = Erol Gelenbe | journal = Journal of Applied Probability | volume = 30 | issue = 3 | date = Sep., 1993 | pages = 742–748 | jstor = 3214781 }}</ref> oroften called a '''Gelenbe network'''<ref>{{cite journal | doi = 10.1016/S0166-5316(02)00127-X | title = G-networks with resets | firstfirst1 = Erol | lastlast1 = Gelenbe | authorlinkauthor-link = Erol Gelenbe | first2 = Jean-Michel | last2 = Fourneau | journal = Performance Evaluation | volume = 49 | issue = 1/4 | pages = 179–191 | year = 2002 }}</ref>) is an open network of G-queues first introduced by [[Erol Gelenbe]] as a model for queueing systems with specific control functions, such as traffic re-routing or traffic destruction, as well as a model for [[neural networks]].<ref>{{cite journal | title = TurningRandom Backneural Timenetworks -with Whatnegative Impactand onpositive Performance?signals |and firstproduct =form Petersolution | lastfirst = HarrisonErol | authorlinklast = Peter G. HarrisonGelenbe | journal = [[TheNeural Computer JournalComputation]] | volume = 53 | doi = 10.1093/comjnl/bxp0211 | issue = 64 | pages = 860502–510 | year = 20091989 }}</ref>| Adoi G-queue= is an [[M/M10.1162/neco.1989.1]] queue with several types of customer:.4.502
| url=https://www.researchgate.net/profile/Erol-Gelenbe-2/publication/239294946_Random_Neural_Networks_with_Negative_and_Positive_Signals_and_Product_Form_Solution/links/60a3b1eb458515952dd4b706/Random-Neural-Networks-with-Negative-and-Positive-Signals-and-Product-Form-Solution.pdf}}</ref><ref>{{cite journal | title = Turning Back Time – What Impact on Performance? | first = Peter | last = Harrison | author-link = Peter G. Harrison | journal = [[The Computer Journal]] | volume = 53 | doi = 10.1093/comjnl/bxp021 | issue = 6 | pages = 860–868 | year = 2009 | citeseerx = 10.1.1.574.9535 }}</ref> A G-queue is a network of queues with several types of novel and useful customers:
 
*''positive'' customers, which arrive from other queues or arrive externally as Poisson arrivals, and obey standard service and routing disciplines as in conventional network models,
*''negative'' customers, which arrive from another queue, or which arrive externally as Poisson arrivals, and remove (or 'kill') customers in a non-empty queue, representing the need to remove traffic when the network is congested, including the removal of "batches" of customers <ref>{{cite journal | doi = 10.2307/3214499 | title = Product-form queueing networks with negative and positive customers | first = Erol | last = Gelenbe | authorlink = Erol Gelenbe | journal = Journal of Applied Probability | volume = 28 | issue = 3 | pages = 656–663 | year = 1991 | jstor = 3214499}}</ref> <ref> {{ cite journal | title = G-Networks with signals and batch removal | first = Erol | last = Gelenbe | authorlinkauthor-link = Erol Gelenbe | journal = Probability in the Engineering and Informational Sciences}| volume = 7 | issue = 3 | pages = 353-342335–342 | year = 1993 | doi=10.1017/s0269964800002953}}</ref> <ref>{{cite journal | doi = 10.1016/S0377-2217(99)00476-2 | title = G-networks: A versatile approach for work removal in queueing networks | first = J.R. | last = Artalejo | journal = European Journal of Operational Research | volume = 126 | issue = 2 | pages = 233–249 | date = Oct., 2000 }} </ref>
*"triggers", which arrive from other queues or from outside the network, and which displace customers and move them to other queues
 
A [[product -form solution]] superficially similar in form to [[Jackson's theorem (queueing theory)|Jackson's theorem]], but which requires the solution of a system of non-linear equations for the traffic flows, exists for the stationary distribution of G-networks while the traffic equations of a G-network are in fact surprisingly non-linear, and the model does not obey partial balance. This broke previous assumptions that partial balance was a necessary condition for a product -form solution. A powerful property of G-networks is that they are universal approximators for continuous and bounded functions, so that they can be used to approximate quite general input-output behaviours.<ref>{{ cite journal | title = Function approximation with spiked random networks | first1 = Erol | last1 = Gelenbe | first2 = Zhi-Hong | last2 = Mao | first3 = Yan | last3 = Da Li | journal = IEEE Transactions on Neural Networks| volume = 10 | number = 1 | pages = 3–9 | year = 1999 | doi=10.1109/72.737488| pmid = 18252498 | citeseerx = 10.1.1.46.7710 }}</ref>
 
==Definition==
{{no footnotes|section|date=February 2012}}
 
A network of ''m'' interconnected queues is a ''G-network'' if
 
# each queue has one server, who serves at rate ''rμ<sub>i</sub>'',
# external arrivals of positive customers or of triggers or resets form [[Poisson processes]] of rate <math>\scriptstyle{\Lambda_i}</math> for positive customers, while triggers and resets, including negative customers, form a Poisson process of rate <math>\scriptstyle{\lambda_i}</math>,
# on completing service a customer moves from queue ''i'' to queue ''j'' as a positive customer with probability <math>\scriptstyle{p_{ij}^{+}}</math>, as a trigger or reset with probability <math>\scriptstyle{p_{ij}^{-}}</math> and departs the network with probability <math>\scriptstyle{d_i}</math>,
# on arrival to a queue, a positive customer acts as usual and increases the queue length by 1,
# on arrival to a queue, the negative customer reduces the length of the queue by some random number (if there is at least one positive customer present at the queue), while a trigger moves a customer probabilistically to another queue and a reset sets the state of the queue to its steady-state if the queue is empty when the reset arrives. All triggers, negative customers and resets disapperdisappear after they have taken their action, so that they are in fact "control" signals in the network,
* note that normal customers leaving a queue can become triggers or resets and negative customers when they visit the next queue.
 
A queue in such a network is known as a '''G-queue'''.
==Stationary distribution theorem==▼
 
▲==Stationary distribution theorem==
Define <math>\scriptstyle{q_i = \lambda^+_i / (r_i + \lambda^-_i)}</math>, where the <math>\scriptstyle{\lambda^+_i, \lambda^-_i}</math> for <math>\scriptstyle{i=1,\ldots,m}</math> satisfy <math>\scriptstyle{\lambda^+_i = \sum_j q_j r_j p^+_{ji} + \Lambda_i}</math> and <math>\scriptstyle{\lambda^-_i = \sum_j q_j r_j p^-_{ji} + \lambda_i}</math>. Then writing <math>\scriptstyle{\mathbf k = (k_1,\ldots,k_m)}</math> for the state of the network (with queue length ''k<sub>i</sub>'' at node ''i''), if a unique non-negative solution <math>\scriptstyle{(\lambda^+_i,\lambda^-_i)}</math> exists to the above equations such that <math>\scriptstyle{q_i<1}</math> for all ''i'' then the stationary probability distribution <math>\scriptstyle{\pi}</math> exists and is given by
 
Define the utilization at each node,
:<math>\pi(k) = \prod_{i=1}^m (1 - q_i)q_i^{k_i}.</math>▼
 
:<math>\rho_i = \frac{\lambda^+_i}{\mu_i + \lambda^-_i}</math>
===Proof===▼
 
where the <math>\scriptstyle{\lambda^+_i, \lambda^-_i}</math> for <math>\scriptstyle{i=1,\ldots,m}</math> satisfy
 
{{NumBlk|:|<math>\lambda^+_i = \sum_j \rho_j \mu_j p^+_{ji} + \Lambda_i \,</math>|{{EquationRef|1}}}}
{{NumBlk|:|<math>\lambda^-_i = \sum_j \rho_j \mu_j p^-_{ji} + \lambda_i. \,</math>|{{EquationRef|2}}}}
 
Then writing (''n''<sub>1</sub>,&nbsp;...&nbsp;,''n''<sub>m</sub>) for the state of the network (with queue length ''n''<sub>''i''</sub> at node ''i''), if a unique non-negative solution <math>\scriptstyle{(\lambda^+_i,\lambda^-_i)}</math> exists to the above equations ({{EquationNote|1}}) and ({{EquationNote|2}}) such that ''ρ''<sub>''i''</sub> for all ''i'' then the stationary probability distribution π exists and is given by
 
▲:<math>\pi(kn_1,n_2,\ldots,n_m) = \prod_{i=1}^m (1 - q_i\rho_i)q_i\rho_i^{k_in_i}.</math>
 
▲===Proof===
{{no footnotes|section|date=February 2012}}
It is sufficient to show <math>\pi</math> satisfies the [[balance equation|global balance equations]] which, quite differently from Jackson networks are non-linear. We note that the model also allows for multiple classes.
 
G-networks have been used in a wide range of applications, including to represent Gene Regulatory Networks, the mix of control and payload in packet networks, neural networks, and the representation of colour images and medical images such as Magnetic Resonance Images.
 
==Response time distribution==
 
The response time is the length of time a customer spends in the system. The response time distribution for a single G-queue is known<ref name="joap-sojourn">{{Cite journal | last1 = Harrison | first1 = P. G. | author-link1 = Peter G. Harrison| last2 = Pitel | first2 = E. | title = Sojourn Times in Single-Server Queues with Negative Customers | journal = Journal of Applied Probability | volume = 30 | issue = 4 | pages = 943–963 | doi = 10.2307/3214524 | jstor = 3214524| year = 1993 }}</ref> where customers are served using a [[first come first served|FCFS]] discipline at rate ''μ'', with positive arrivals at rate ''λ''<sup>+</sup> and negative arrivals at rate ''λ''<sup>−</sup> which kill customers from the end of the queue. The [[Laplace transform]] of response time distribution in this situation is<ref name="joap-sojourn" /><ref name="net-resp" />
:<math>W^\ast(s) = \frac{\mu(1-\rho)}{\lambda^+}\frac{s+\lambda+\mu(1-\rho)-\sqrt{[s+\lambda+\mu(1-\rho)]^2-4\lambda^+\lambda^-}}{\lambda^--\lambda^+-\mu(1-\rho)-s+\sqrt{[s+\lambda+\mu(1-\rho)]^2-4\lambda^+\lambda^-}}</math>
where ''λ''&nbsp;=&nbsp;''λ''<sup>+</sup>&nbsp;+&nbsp;''λ''<sup>−</sup> and ''ρ''&nbsp;=&nbsp;''λ''<sup>+</sup>/(''λ''<sup>−</sup>&nbsp;+&nbsp;''μ''), requiring ''ρ''&nbsp;<&nbsp;1 for stability.
 
The response time for a tandem pair of G-queues (where customers who finish service at the first node immediately move to the second, then leave the network) is also known, and it is thought extensions to larger networks will be intractable.<ref name="net-resp">{{cite conference|title=Response times in G-nets|conference=13th International Symposium on Computer and Information Sciences (ISCIS 1998)|isbn=9051994052|pages=9–16|first=Peter G.|last=Harrison|year=1998|author-link=Peter G. Harrison|url=https://books.google.com/books?id=ClU2XG64UOIC&pg=PA9}}</ref>
 
==References==
{{Reflist}}
 
{{Queueing theory}}
[[Category:{{Stochastic processes]]}}
 
[[Category:Queueing theory]]
Retrieved from "https://en.wikipedia.org/wiki/G-network"