• 検索結果がありません。

ON A QUEUEING SYSTEM M/G/m/(m) HAVING THE PRIORITY ORDER AMONG SERVICE STATIONS

N/A
N/A
Protected

Academic year: 2021

シェア "ON A QUEUEING SYSTEM M/G/m/(m) HAVING THE PRIORITY ORDER AMONG SERVICE STATIONS"

Copied!
6
0
0

読み込み中.... (全文を見る)

全文

(1)

oN A QuEuEING sYsTEM M7G/m(m)HAvlNG THE

  PRIORITY ORDER AMONG SERVICE STATIONS

       BY

YoslRo TUMURA

  1.Consider the many servers system where the arrival is M, and in every servi㏄ station a customer is served according to the same distribution function B(∫). It is assumed, moreover, that a customer丘nding at his anival all stations being busy must leave the system, and a customer finding some stations idle enters one of them 。at random with equal probabilitγ. Let us call such system‘‘random case’, in brief. B.A. Sevastyanov[1]has proved:   If 3(s)has a五nite mean 1/μ, then the systemルηG/m(m)has the fbllowing sta− ’tionary distribution:       MI        ・ 1}1/Σ6(一・・1・……)        ツ=0 −whereρ=2/μ,λbeing an arrival rate・   That is:the equilibrium probabiHtiesρ。 depend only onρand m, but not on the ”distribution B(s).   In some practical problems(e.g. telephone exchange)aIl stations are not symmetric 『but ordered so that a customer who arrives at a system is decided(in deterministic) ・to enter the station having the smallest number among idle stations. Or, to every モ盾高b奄獅≠狽奄盾氏iξ1,…,ξ.;1≦ξ《≦〃,〃≦〃1)correspond the probabilit▲es, such that:a .customer who丘nds at his arriva1ξ‘−th stations(i=1,…, n)being idle and all others ’busy enters one of idle stations at random with the above de丘ned probabiHty.   Let us consider, i皿this note, the fbrmer case, which we shall call ip brief‘‘priority .ease.,,@It is in question to find not only equi且brium probabihtiesρπthat n stat三〇ns are.busy, but also probab狙tiesπヵthat n−th station is busy. This last problem is ’usefUI in practice, fbr example, to deal with the reliability of a machine system.   We have proved in the previous paper the genera1 theorems of queueing theory[2], −which we shall apply to our system.   Let n(レ1,… ,レヵ)(〃≦〃z)be a state thatソ1, ソ2, …, レバth stations are busy and .all others idle(assumingレ1<レ2<…<レ。). B(s)is assumed to be absolutely contin. uous ex㏄pt at most{sy}where lim sv = oo. The equilibrium equations fbr such sys一 託em are aS f()llOWS;

(1)・p・一

ー1…(・・)S.iglizlBSsfl.)・ [60]

(2)

(2)   (3)

where

and(2)and(3)hold except atぷ5=・ぶ夕. 1・… ツカare not placed in ascendhlg ord (4) (5) 互fthe sequence{ぷγ}is not empty, We add the conditions:   孔(〃、,…,〃。)(ぷ・、,   Our general theorem shows that if(1)∼(5)have『non−negative solution po>0, then the system has the equinbrium. state, 血e state η(Vl,’”, Vn)

      ON A QUEUEING SYSTEM M/G/m(m)HAvlNG        61

 fi [Σ(、書。、+・(・))+・]Pn・……・・。・一Σ!Pn・・[一〃.・コ(・〃・・…・・㌦…)1竺麗i)  1      ξキyi        (n=11,2,「… ,〃z−1; 1≦ッ1<.V2<’… <〃外≦〃1), th Σ(£+・(・・))Pm(……・・鋤)一・・  1       ・(・)一際ぷ)/[1一β(・)コ,       The notationπ[〃1)…,〃ヵ]shows that〃1,       er. Thgs6 equations have boundaτy conditions:    P1(1)(0)= Zpo, P1{i)(0)=0. .(i=2,・.・・,〃1), P・・………(・〃・・…・・〃・一・・・…’……・s・P−{;孔捲穿一・・ ’’”・・)’f”’=i         (n =2, 3, … ,〃2; 1≦び1<〃2… <〃”≦〃1; Z=1, … ,n).      …・s・・一…+・・s・1・ピ…s・2−}三』瓢恥・…(1・・1 s・−d,…)        (〃−1,….,〃1;i=1,…,〃).        such that        and the eqUilibrium probability of       is defined by      .    ・  .    .         ・  . ・  ・        〃・・一・一!・…、・…艶・・ Hence, we have to prove, at first,ρo>0......   2.It is not easy to solve(1)∼(5)directly, therefbre we shall prove some prop− ¢rti・・indi・㏄tly;.狽?Et i・,・・mp・ri・g wlth㎞・㎜・e・ult・.   Let      .        が        ρ1ω一ΣP1ωω,       ‘=1 ・then

(1’)   ・p・−1・・ω1竺甥,)

rls satis耳ed丘om(1), and from(4)we have   (4’)       (∼1(0)==ZPo.』   Next, fbr(’1,τ2,…,’ヵ;η≦m), there eXist n!permlltations(’ξ1,…,’e.). Consider 芭he mean:      ,、   、        ・・… ・(’・・…・’・)r:,ΣPn・。・一・(’・・・・…t・・、)(剛・

       @・(’・・…・t・・)−i,ΣPm(’……・・⊇・.・    ’

・ .

(3)

62

Y.TUMURA

and the辻s㎜

       ρ・(’・・’”・t・)言、,顯・・……・…(’・・’”・t・)ぴくt”)

w・・忙・h・・一・i・…ad・・d・ve・a・(m)・・m・・na・i・n・・・……・・・・…−

have from(2)and(3)respectively       #

(・’)   [Σ(sl、+・ω)+・]ρ・一・

       ’=1

and

      Pt

(・’)   Σ(募、+・ω)ρ・一・・

       1       ’   ’As boundary conditiOns we have also from(5) (・’)0。(’・,…,tt’・一・,・,’ ,…,・・)一÷ρ・一・(’・,…,’・一・,’…,…, t・)       (π=2,… ,〃1; i=1, ・・.・,π). These equations(1’)∼(5’)are equ丑iblium equations fbr a random case. From defini− tions of〈2π, it is dear that       ,.、頁墨…,一・−1・・(’・・…・輌・ ’Hen㏄, from Sevastyanov’s theolem, or solving(1’)∼(5’)directly we have   THEOREM 1. For the』ツぷ’θ〃1ルr/G/m(m)with prioritγぷtatio〃, the eguilibliu〃1 Plrob− ability吻’ノus’nぷta’ions oアe busy is equal’o       Pt

(・) ・・≡Σ蜘 ・一£/r/Σ葺{(・一・・1・…・m)・

      (〃1,…,Pes)       〃=o and the〃lea〃queueぷize is given∂y        が      m       ぐ

(・)  ・≡Σ・・ ・一篇1/Σel・

       兄=1       0   Clearly po>0, so that allτesults are ensured by our general theorem.   3. SupPose〃1<〃2<… <〃外く〃2, and let us write

(・)  P…一・・−Pn・・。・・…+!・・・…一轟

       ぴ一1.…,m−1;1≦v、<v,、<…<〃。≦〃1−1)・ For n≦n卜2, i皿tegrating byぷ”the equation that is inserted Pn+1(夕1,_, v4,伽}in(2), we

have

      # (・L)…hand・・d・一∫[Σ(、隻、+・(…))+、i。+・(spt)+・]Pn…〃… 轟       ;”1        −[Σ(、竃、+・(…)+・)!?”+・……・一・d・・       1

(4)

      ON A QUEUEING SYSTEM』胡G1〃1(m)HAVING        63       +!・・・・….…。・・1蒜;)・ because of       Pn.、、。、,…,㌦,。,(・“、,…,・“。,0)−0,   ’ ・

and

(・R) R・gh・hand・・d・一Σ1[1・・・・….・・ 嗣1雲曇)・

      ξキ〃∠,餌 Inserting P鈴(〃1,…,〃扮)in(2)we have (1・)[Σ(、i。、+・(…))+・]…〃。−P        一Σ∫Pn・・[       ばβ(ぷξ)”・ジ”・”孤・ξコ1一β(ぷξ)+∼Pn… ・ 一・1≦鵬㍑)・       ξキ〃i,術 Adding(9)and(10), and considering the de丘nitio11(8)』       り (・〃)[Σ(、皇、+・(…))†・]pX・,…,・・一Σ1嚥・[・、,・・_・・1弩農)       ’=1       ξキ〃’    馳       ξ≦伽一1       (〃=1,… ,〃1−2; 1≦ジ1<〃2<・二・<γヵ≦〃1−1). Forπ=7η一1, integrating(3)byぷ邪, and from(5)we have also        m          Σ(£、+μω)!Pmd・.+∼・㌦1≦至裂:…)一・・       1 Adding(2)inserted Pm−1(1,_,が一1}to this equation we have        づ      (3〃)         Σ(£、+μ(ぷ・))P‡一・一・・       i=1 1nserting P1ωin(2)and hltegratillg.1)yぷ傷we have, as P1(邪)(0)=0,        m         ・!…−1盈ω1霊)一Σ![P・⑭d・・コ1弩農)・       1 Adding(1)t・thi、 equ。ti。亘w。 hav。       mロ 

(1”)   ・・9一Σ1・f(・1弩{曇)・

      ’=1

where

      ・ぎ一・・+!P・・轟 (1”),(2”)and(3”)ar6 the eqUilibrium equations for M/G/m−1(m−1)with priority stations. As boundary c皿ditions we hav?easily from(4)and’(5), P数・・(・)一{『ま:1≡;, ・f・  ・(ぷ〃、,…,5〃,.、,0,s・i.、,…,&。)一{;Pエ1( ’・  ”㌔弓1:;: Thus, we have

(5)

Y.TUMURA

  LEMMA. Letク。(。1,_,レ孤)and p#tv1,_,“”}be eqpilibrium.probabilities fbr the systems. M/G/m(m)andハイ/G/m−1(m−1)with priority Stations・ respectively・ Then, the prob− abiity that i−th station is busy is, fbr such systems,      ’       楊      ’    溺一1       πi一Σ  Σ ρ。[・、,….・。一、,i]=Σ  Σ P‡[〃、,…,〃。.、,‘]・        外=1(〃1,.・・,レ外一1  レξ≦m)      カ=1(〃峯梁了1)   Now we shall prove   THEOREM 2. For theのノぷ’θ〃1 M∫G/m(m)with prioriリノぷtationぷ,.the equilibriu〃i prob− 4b助γπ.吻’左イ乃ぷtation is busy is equal to        k_1       .k

(11) …(芸1),/Σ芸一瑞1/Σ芸(k−1……・m)・

      ”       ..       0       0 whereρ=λ/μ. PRooF. For k=1, from the Lemnia and Sevastyanov’s theorem伽〒1),       π1=ρ/(1十ρ) is evident. Co皿sider the systems M/(}/n(n)and M/G/n−1(n−1). For the fbrmer, the mean queue size is given by・一     一   ・−        n        L一Σπ5.       ‘=1 From(7),        n      れ       Σ・i−・一芸三/Σei・        1       0 π克 .7 6 .5 4 , 3 . 2 . 1 . ユ       eぶ        &     ざ 烏

⇔  急

2 3 4    5    6    7    0rder of station    ’ Figure. Values ofπk 8 ・㌧ 9 1 10

(6)

       ON A QUEUEING SYSTEM 1吻(7/m(〃i) HAVING ’    iFor the later system       ”−エ       n

.    Σ・・†(ii・−lll”,),/Σ;i・ 、

       」       1      0    whereπ∼s for both systems are the same. This fact Proves the Theorem.      We illustrate the values of zk.fbr:variouSρ,s. 65

REFERENCES

’[1] 匡2] Sevastyanov, B. A.(1957):An ergodic theorem for Markov processes and its applica・  tion to .telephone systems with refusals. Theor. Prob. App1.,2,104−112.・ Tumura, Y.(1967):Equilib血m equations methOd in generalized queueing problems.  TRU Math.,3,48−61.      .      ・

RETRACTION

YOSIRO TUMURA

  In the previous paper(Equilib亘um equations method血the generalized queueing proレ 1em. TRU Math.,3), a counter exalnple to Finch,s result(Acta Math. Sci. ACad. Hu耳g.8) has been shown, and Fmch’s incorrect statement on Kiefer and WolfOwitZ’s result(A. M. S.,78)was copied. Kiefer and Wolfbwitz did not state, however,‘‘0−state’, is recurrent, but devoted themselves to getting around the di伍culty that O−state n㏄ed not be recurrent. The authur wishes to retract olle lhle(1.26, p.51)and co仕ect it as fbllows:   .   Tlle fbllow短g example, however, throws d皿bt on Finch’sτesult.  , ノ .

参照

関連したドキュメント

The main purpose of the present paper is a development of the fibering method of Pohozaev [17] for the investigation of the inhomogeneous Neumann boundary value problems

Zhao, “Haar wavelet operational matrix of fractional order integration and its applications in solving the fractional order differential equations,” Applied Mathematics and

We initiate the investigation of a stochastic system of evolution partial differential equations modelling the turbulent flows of a second grade fluid filling a bounded domain of R

Also, extended F-expansion method showed that soliton solutions and triangular periodic solutions can be established as the limits of Jacobi doubly periodic wave solutions.. When m →

Since all shift morphisms are bounded sliding block codes in the finite alphabet case, this implies that if K is any field, and if E and F are finite graphs with no sinks and

Dimension Type of theory Boundary/corner structure Invariants 4k Topological gauge theory manifold with corners of codim-2 primary classes 4k − 1 Chern–Simons theory boundary

If we represent π by a diagram (of either type), erase the point corresponding to i and the arc connected to the point (and number other points appropriately for the circular

This shows that although each group A n is algebraically compact (it embeds as a pure subgroup of the compact group Π ∞ 1 (Z/2)), the structural maps are not continuous (in the sense