ON AN EVALUATION OF C且ARACTERISTIC BEHAVIOR
CONCERNED WITH TANDEM QUEUING PROCESS
BY
TOJI MAKINO
1.亘ntroduction. The Tandem type queuing system is, from the practical point of view, one of the㎞portant topics in the theory of queues. However, we can not say that we have enough results ill this丘eld to apply them to individual prac− tical problems. The useful techniques introduced by Lindley[1]have bee!l apPlied to the study of Tandem type systems as wen as other various types of queui皿g systems, and have produced interesting works by Reich[7], Sacks[12], and recently, papers by Ghosal [16],Mastersoll and Sherman[17]. Jackson[3],[4]used another technique and obtained the distribution of length of waiting line at each stage or service station of a multi−stage Tandem type queuing system. III all of these papers are treated the cases where no restriction is made on the length of queues at each stage of queuing systems. On the contrary, there are not so many studies in the case of the limited size of Waiting rooms at each stage..The・reason why is the di伍culty of analysis which due to the o㏄urren㏄of blocking effect. In thiS COntext, Hunt[5]Obtained the va1Ue of maxirnum posSible utilizatiOnρmax fbr some types of Tandem queue, and Finch[9]fbund the distributidn of waiting t㎞ein the system with waiting roolns. In[9]was treated the single stage system with limited size 2V of waiting room. Finch mainly discussed the type M/M/1(N) in detail. His result was extended by Jain[15]to the typeルt/E,/1(N)and Jain presented the table of customer lost fbr various values of utilizationρand various size 2V of waiting. room. Kishi[13]a皿d Makino[18], in conn㏄tion with Hunt[5], discussed two−stage Tandem system with Poisson arrival distribution and exponential service distribution. The results of these papers were extended by Makino[19]to the tvyg−stage systenl where the number of first stage station is larger than one. In these works was used the method of g皿erating functions. However, using such adevice the study of more complex systems will be rather di{ficult. To avoid this di伍culty, introducing the concept of Mean Passage. T㎞e, Makino treated two and. three srage Tandem systems in his paper;18
T.MAKINO
‘‘On the Mean Passage Time concerning some queuing problems of the Tandem 寸ype.” [21] Shortly after the recept輌on of[21]by OR Society, Suzuki published‘‘On a Tan− dem Queue with Blocking.”[20] These two papers published independently, and the fbrmer is practical, the latter .is rather theoretica1, but the themes treated are quite similar. The main purpose ・of[20]is the study of blocking eff㏄t which is similar to the study treated in[21]. However the model treated in[20]is the two−stage system without the waiting room −at the s㏄ond stage. Since blocking ef£㏄t is influenced by the size of wait輌ng room at the second stage, the study of the system with waiting room seems to be necessary. But by the method of[20], the analysis is diMcult because of the increase of the .number of states. Of course, we do Ilot n㏄d to consider blocking effect in the case where infinite ・queue length is allowed at each stage. The problem of blocking occurs only fbr .the system with limited queue length. The purpose of this paper is to present the practical table of mean blocking rate .PB fbr each size of waitillg room at each stage of the system. The systems treated.here are those of three stages, each stage havillg s姐gle service station, With exponential service time distribution, and the&st stage station being ・never empty. Noting that [M…Passag・T・m・コr三馬去・ (μ1:mean service rate at the丘rst service station) −we may investigate[M.P.T.]in place of PB. We know teat the reversibility of[M.P.T.]holds fbr the systems in[21]. This property which hold fbr血ore extended class of systems is very usefu1 for numerical ・calculation of PB. Moreover, treating usual system(considering the arrival customers皿ake the queue :in front of the first service station), the necessary and su丘icient condition that the system be in steady state is obta輌ned dir㏄tly using the value of[M.P.T.]. Con− ・ceming this point, we shan discuss it in paragraph 4. 2.Balance Equations. We are going to evaluate some characteristics of the sys− ^tem with the structure model of following chart. (Warehouseeirst queue)⇒ 1 First stage serVlce S楓tlon →Second
queue
→ S・。・ndil stage i→: serVlce ・ statlon (servlcerate;μ1)Third
que「e →Th辻d
S伍ge servlce statlon ご麗9)(serVlcerate;μ2) Fig. 1 (waltlngroom)⇒Output
(servlcerate;μ3)ON AN EVALUATION OF CHARACTERISTIC BEHAVIOR CONCERNED
19 SupPose that in the s㏄ond queue and the third queue are anowed finite number of customers M, N, whereas in the first queue exist always infinite number of cus− tomers, i.e. assume’that there are a warehouse instead of丘rst queue. The service time distributtons at the first, s㏄ond and third station are assumed to be exponential with service rateμ1,μ2 a皿d・μ3 respectively.. The symbol伽, n)of the state represent the fbllowing situations;the secOnd and the third station are in busy, moreover the queue lellgth of secolld and third queue are(m−1),(n−1), resp㏄tively. The steady state probability that the system be in the state(m, n)is denoted by P。,n. Then we have the following balance equations. State (0,0) (0,1) (0,2) (0,N) (0,N+1) (0,N+2) lNo. of cus. tomers lst at the ・t・・i・n認. 1ng l rOO皿2nd
lstati°n110
1 1 1 1 ’ 0 0 0 0 0 1 (1,0) 1 (2,0) ⋮: (κ,0) ⋮: (M,0) (M+1,0) (M+2,0) −一丁− 0 0 0 0 0B
11一−一β
0一1
k−1ゴ
M
÷
−一− 1 −一−一−上 No. of cus. tomers 3rd at the 3rd station麟
0 0゜11
1 」V−1 1 1N
1N
1 〇一〇 00一〇︸0
0[0
︵UO∨0
(for 1≦9m≦二M,1≦:n≦9N) (m,n) 1 m−1 1 n−1 1 Balance Equation (where,μ2/μ1=γ2,μ3/μ1=γ3) γ3Po1−Poo=0 γ2Plo+γ3Po2−(1+γ3)Pb1=0 γ2P11+γ3Po3−(1+γ3)Po2=0 γ2Pl,N−1+γ3Po.N+1−(1+γ3)Po,N=0 γ2P1,N十γ3Po, N+2−(1十γ3)Po,N+1=0 γ2P1,N+1−(1十γ3)1㌔,N+2=0 Pb,o十γ3Pll−(1十γ2)PiO=O PI,o+γ3P21−(1+γ2)P20=0 Pk−1,0十γ3Pk,1−(1十γ2)Pb,o=0 PM−1,0+γ3PM,1−(1+γ2)PM, o=O PM, o+γ3PM+1,1−(1+γ2)PM+1.o=0 ’』\ PM+1,0+γ3PM+2,1一γ2PM+2,0=0 Pm−1,π+γ2」㌦+1,蕗一1+γ3P.,カ+1 −(1+γ2+γ3)Pm,銘=020
T.MAKINO
State 1st station (M,N+1) 1 (M+1,N) 1 (M+1,1V+1)11 t (M+2,1) (M+2,2) (M十2,N十1) (M十2,∧r十2) .. ’B
B
B
B
No. of cus. tomers at the2nd
wait− ingroom
M−1
M
M
2nd
stationL鑑.
tomers at the 3rd wait・ ingroom
11N
1M
1 1M
1M
1M
B
N−1
0 1N
3rd station 1 1 1 Balance Equation (where,μ2/μ1=γ2,μ3/μ1=γ3) PM−1,N+1+γ2PM+1,N+γ3PM, N+2 −(1+γ2+γ3)九,肝1=0 PM. N+γ2PM・2,N−1+γ3PM+1,N+1 −(1+γ2+γ3)PM・1,N=0 PM,N+1+γ2PM+2,N+γ3PM+2,N+2 i −(1+γ2+γ3)PM+1,Ntl=O 1 1 1 1 PM+1,1+γ3PM+2,2−(γ2+γ3)PM+2.1=0 PM+1,2+γ3PM・2,3−(γ2+γ3)PM+2,2=O PM+1,N+1」(γ2十γ3)PM+2, N+1=0 PM,N+2.+γ2PM・2,N+1一γ3PM・2,N+2=0 (1, ハr十2) 1 | | 1B
N
1 Pb,N+2+γ2P2,N+1−(1+γ3)Pl.N+2=0 (M,N+2) 1M
B
1 PM−1,rv「+2十γ2PM+1,N+1−(1十γ3)PM,ハr+2=0 Where the symbols O,1,8denote the states that the service station be in empty, in busy, and in blocking, respectively. We want to丘nd the value of」Piゴ. For this purpose we have take care that the normality condition holds. That is, we can obtain Pii by noting that ΣP已一1. ‘,5 Thus we can calculate the mean blocking rate PB. FurthCrmore the mean length of the second queue and the third queue will be obtained by fbllowillg procedure; First, denote by X and Y the.queue length of the second queue(second waiting roo皿)and the third queue・(third waiting room), respectively. Let us denote by E(X), E(Y), E(X十γ)the Expectation of X,〕陥(X十y). Then, we have M ユ M ユ N E(X)一Σん・{Pk.、.。+Pk,N.2}+ΣΣ{(〃1−1)・Pm,外}+MΣ1「M.2.. k=1 n=1 秘=1 カ=J E(・)−狽s,k・{P・, +P…,…}+螢{(−1働+N{誉九,…+PM・ ・・}・
P.一Σ PM .2,k. E(x+Y)−E(x)+E(Y). k=O Where M and 2V are limitations of the size of the second and third waiting room. The numerical value of above expression in the case of μ1=μ2=μ3 .ON AN EVALUATION OF CHARACTERISTIC BEHAVIOR CONCERNED 21
will be shown ill Table 1. Table L State Probabilities, Expectations of the Queue Length and M剛Blocking Rates in t血e Case ofμ宜=μ2=μ3M=0,Nニ0
・ta・・1oo
●01
02
10
20
112122
E(x) E(y) E(x +Y)PB
Prob. 0.1026 0.1026 0.0769 0.1282 0.2051 0.1539 0.0769 0.1539 0 0 0 0.4359 ルf=0,2V=1 ・t・t・10123000001
01212
21ふ122
23 E(x) E(Y) E(x +Y)PB
Prob. 0.0933 0.0933 0.0800 0.0400 0.1067 0.1867 0.1200 0.0800 0.0800 0.0400 0.0800 0 0.320 0.320 0.3867M=0,N=2
State 00 O1 O2 O3、
00123
ープ一11▲−
12今34﹁
∩∠︹∠︵∠2
E(x) E(Y) E(x +y)PB
1…b・
0.0879 0.0879 0.0769 0.0440 0.0220 0.0989 0.1758 0.1099 0.0659 0.0440 0.0769 0.0440 0.0220 0.0440 0 0.538 0.538 0.3626lMto,N=3
1・t・t・1馳. 00 O1 O2 O3O4
5001∩∠
01︵∠111
︹541▲23
11︵∠22
45
ラ一2
E(x) E(y) E(x +y)PB
0.0849 0.0849 0.0745 0.0435 0.0248 0.0124 0.0952 0.1698 0.1056 0.0621 0.0373 0.0248 0.0745 0.0435 0.0248 0.0124 0.0248 0 0.689 0.689 0.3499 M=1,ハr=0 State 00 O1 O2 P0チ
0111う一
︵51︵∠31
32 E(x) E( Y)E(x
+Y)PB
1 Pr… 0.0533 0.0533 0.0400 0.0667 0.1000 0.1667 0.0800 0.1333 0.0667 0.0867 0.1533 0.707 0 0.707 0.3867 ハ∬=1,N=1 State 00 O1 O2 O3 P0︵UO121▲
︵∠31ーエ︵∠
︵∠1233
23331
E(x) E(γ) E(x +y)PB
1蹴・
0.0539 0.0539 0.0485 0.0322 0.0594 0.0759 0.1403 0.0648 0.0645 0.0925 0.0725 0.0644 0.0362 0.0886 0.0524 0.623 0.395 1.018 0.329522
T.MAKINO
M−1,N−・IM−1,N−・
State i iIM=2・
N=0
…b・1・・…1…b・1・ta・・1・・… M=2,∧「=1 State Prob.M=2,N=2
・・…1 Prob. M=2,N= 3 State l Prob.Ol︹∠︵﹂41
00000
1 0.0518i OO l O.05181 01 : 0.04851 021:鵠92
10 0.0551 05 20 0.0671 10 30 0.1256. 20 … 1;i9:1:12…i?31﹁∠31
12223
0.0416 1 0.0791i l O.0533i l O・0449i ゜・°585自∠ぺ34丁12
111︹∠へ∠
0.03791 23 32 33}。.。225i24 1 34 0.05531 31 14 0.0328i 32 … :33 … E(x)1 E(Y)Eζη1
PB
0.577 0.709 1.286 0.2997455
3つ⊃−
i i l i : 1 … E(x)l i E(y),xi
E( +γ). PB l i 1 0.04ggi O.0498: 0.0471i l O.0392. 1 0.02571 ・.・1351 0.05241 1 0.062si O.1172‘ O・0551i i O.0527i E O.03781 0.0271i ; O・07311 0.0471… 0.0342 0.02911 0.0545… ! 0.0358i ; 0.02441 i O.0146’ 0.0359 0.0213・ 0.550 O.955 1.505 0.2822 00 O1 O2 P0 Q000111
惚﹂4112︵﹂
−︹∠︹∠﹁∠ 414﹁−︵∠1
E(x) E(Y) ・1㍍)lPB
i O.02931 1 0.02931 0.02201 0・0366} 0.0550; i O.08971 1 0.1520i O.0440i l O.0733i O.1245i O.0623i O.1484i O.0476: 0.08611 1.502 0 1.502 0.36260123000001
00012
2341玉
−︹∠121
︹∠︹∠3341
︹∠333
44﹁イー︵∠ IE(x)l lE(Y) E(x +γ)PB
。.。。631 O.0034・ 0.0034, O.00301 i O.002L l O.0037i l | 0.0045i i O.12051 | 0.04021 1 0.04141 1 E O.0540i O.0587i O.0808・ 0.0680; 0.0574’ d ・…4・1 0.0878 0.0397 0.0538 i 1.174 O.388 1.562 0.299701∩∠34丁
00000
0000112341
23123
11∩∠2︵∠
−︹∠312
3334.4
3444
44﹁イー2
; 0.0340: … 0.0340. 0.0323 0.02721 0.0171i i d O.03571 0.0408 0.0529 0.1026. 0.0374、 0.0374i O.03411 0.0458・ L。.。471i O.0389: i ︶ ︵E
i E(Y)lEα1
+γ)lPB
0.0651 0・0468 O.0436i O.04971 i O.0343i l i O.02工8; ) 0.0576i シ 0.028ぴ … 0.0358i i … i 00 , 01 02 03 04’︶000nU
O1234﹁
12341
1111︹∠
23412
2223つ﹂
3412333444
44i ll目 i251 1 i l 1.。,6i。(X)} … 1 ’E(Y)t O。806 ;: ::kl㌣︶i
0.0332 0.0332 0.0320 0.0286 0.0221 ■ 0.0124 0.0343 0.0380 0.0479 0.0923 0.0355 0。0352 0。0317 0.0248 0.0417 0.0418 0.0326 0.0269 0.0577 0.0392 0.0312 0.0299 0.0442 0.0311 0.0231 0.0150 0.0398 0.Ol97 0.0248 1.153 1.120 2.273 0.2457ON AN EVALUATION OF CHARACTERISTIC BEHAVIOR CONCERNED
23
M=3,ハr=0 State 00 O1 O2 P0S
30I
011
<>−▲2 11些ーユ22 34τ5ーエ角∠ 2︵∠ 3く∨ E(x) E(y) E(x+Y)PB
Ik・b・
0.0166 0.0166 0.0124 0.0207 0.0311 0.0507 0.0854 0.1455・ 0.0248 0.0414 0.0704 0.1201 0.0600 0.0269 0.0487 0.0844 0.1444 2.358 0 2.358 0.3499 M・3,」V=1・t・t・1…b・
00 O1フ0310
M=3,N=2
00001
2︹34﹁51
State 887”︶0/l10/33
2211200000
ロ の コ コ000︵UO
1⑪:鍋
00 O1O2
O3 O400000
ーエ︵∠34﹁5 1 …b・ 0.0235 0.0235 0.0224 0.0191 0.0123M=3,N=3
12 、21 22 31 3212412⇔544二︶55
0.0571i O.1108i O・0260 i I O.026gl O.0348i O.0390’ 0.0492 0.0556 0.0749 0.0649 0.0537i l:器}引12312
111︹∠2
⑲⊃1231
23334
13 0.0262 42 23 0.04091 43 33 0.0529 1 51 1 52 53 E(.¥) E(Y) E(x+γ)PB
i
2.095 0.457 2.552 0.28224444
㌘﹂−ふ23 E(x) E(y) E(x+Y)PB
State 00 O1 O2 O3 O41
Prob. 0.0246 0.02761 0.0336 1 0.04571 l .O.09071 i : 0.02571: O.0259i O.02471 0.0306; 0・03261 1 … O.03231⑪:麟1
0.03711 : 0.0578| 1 … 1:腸;i ⑪:8111i O.0210i 。.。569 I O.0223i O.0297i O.03591 … 1 1.922 0.866 2.788 0.2457 | ︷|
0518
281
Ol∩∠34
ξ﹂111A111234122223
︹∠34丁−︵∠33344
34123
44﹁く∨55 4丁5∼︾r︶’︶55123
E(x) E(Y) E(x+γ)PB
|1
i
0.0236 0.0236 0.0229・ 0.0209 0.017b 0.OIO3・ 0.024Z O.026Z O.0305 0.0400・ 0.0789・ 0.0249・ 0.0249・ 0.0236 0.0207 0.028Z O.0292. 0.0279 0.0240・ 0.034ア 0.0361 0.0296 0.0263 0.0494・ 0.0347 0.0292、 O. 0296 0.0389 0.0284 0.0220 0.0148、 0.0404 0.0172. 0.02171 0.0257 1.806 1.226 3.032 0.2233:24
T.MAKINO
3.Some E▼aluations on the Numerical Values of the Mean Blocking ’ Rate PB .in the Case ofμi=μS2=11e. Using Table 1, we get the value of(1−PB). These values will be shown in the Table 2. Let us denote by」PB(M, N)the meall blocking rate in the case where the size of :second waiting room is 1imited by N. Then we s㏄that PB(M, N)−PB(凡ルの ’using the Table 2. Table 2. Numerical Value of eB(=1−PB) (forμ、=陶=k) 0 0 10.5641 1 2 3 0.6133 0.6374 1 4 5 6 0.6501 O.6572 0.6612i
0.6635 ∞ 2/3 1 0.6133 0.6705 0.7003 0.7178 0.7286 0.7359 3/4 2 0.6374 0.7003 0.7340 0.7543 0.7676 3 4 0.6501 0.7178 0.7543 0.7767 0.6572 0.7286 0.7676 4/5 5/6 6/7 5 0.6612 0.7359 7/8 6 0.6635 8/9 oo 2/3 3/4 4/5 5/6 6/7 7/8 8/9 1 The fact pointed out above may also be clear by the argument that the revers− :ibility of[M.P.T.(Mean Passage Time)]holds. To evaluate the effectiveness of the waiting room, it is convenient to calculate the achievement rates. Where the achievement rates are de丘11ed as fbllows; A≡孟1吉1麗吉吉鵠(8)も)} M,ハr→。。 22 {1−P8(M,1V)}− 39 17 , 万 ・・≡li量1譜謬2吉甥1。1)1)} M→。。 22 {1−P.(M,N)}− 39 1V十2 22 ’ 1V十3 39ON AN EVALUATION OF CHARACTERISTIC BEHAVIOR CONCERNED
2∫ ん≡1、蓋1言i耀識吉≧碧1。1)1)} 」v→。。 22 {1一PB(M, N)}− 39 M十2 22 ’ M十3 39 The numerical values of achievement rates are shown in the Table 3. Now Let us discuss the property of the system using the numerical values in the・ preceding tables. At{irst, consider the probability Q・(M,N)≡1−PB(M, N), that the first station be not in the blocking state. From the property of the system, we see easily that the follOwing formulas hold:・・(…)一誇…(…N)一芸i…(M,・・)一鵠
Now, we shall compare the value of(2.(M,1V)with above values, notillg that二 the value of eB(M,1V)is monotone increasing with the values of M and 2V. By using the Table 2 we see, fbr example, (∼B(0,4)=0.6572. Table 3. i) Numerical Value of AN Achie▼ement Rate(%) (forμi=二μ2=k) ’、DxN
\ 0 1 20123456
11
1
00.0 00.0 00.0 00.0 00.0 00.0 00.0 48.0 41.8 38.7 37.0 35.7 34.5 71.5 63.6 59.6 56.9 55.2 3 4 「 ・ 90.8i 84.3
i80・1
1
5 83.9 76.4 71.9 69.1 94.7 92.6 6 96.9 ii)Numerical Value of AM 00123456
00.0 48.0 71.5 83.9 90.8 94.7 96.9 11
00.0 41.8 63.6 76.4 84.3 92.6 2 .| ︽ 00.0 38.7 59.6 71.9 80.1 3 00.0 37.0 56.9 69.1 4 OO.0 35.7 55.2 5 6 。。.。1 34.5 00.026 T.MAKINO
This value is close to the value of 2 ρ・(o・°°)=了・ In other wards, we s◎e the achievement rate b㏄omes go.8%. Therefbre, in practice, we can conclude that if M=O, 1V≧4 寸hen the system may be treated that with the third station which is assumed to be .an independent channel. 〆 On the other hand, f()r fixed M(or 2V)the values of .警)(・・¥/x)) ’ ・ べwill also be a measure of the effect of the waiting room. . 4.Ergodicity of a System. The necessary and sumcient condition that the sys一 寸em be in steady state is discussed very often. So, we shall consider the relation between the condition of ergodicity and[M.T.P.]. At first we are going to study the ergodicity for the system with two phases which ゴssimilar to the system treated in the pr㏄eding paragraph. That is, .『 1 .一一・− 1 ;i i穰e鵬:⇒・・rs・・ta…n→・・…nd q・…i→・・・…d・ta・…!⇒・咽
(wa輌ting room) Fig・1 We have already calculated the[M.P.T.]ρn the above syst㎝. III the study of the system with above structure, we supposed the probability that the且rst stage be in empty to vanish. ・ But we usUally treat the system with the following model; ; −1 【 i l I・p・・⇒w一し竺・……n]→じ…ndq・・u・−i・・c・nd諏…ni⇒…p・・
qUeUe (Waiting rOOm) Fig. 3 That is, if a customer arrive to the system at the time where the丘rst stage is in busy(or block), then the cuStome士have to wait in the丘rst queue. Where the ]ength of the丘rst queue may be infinite. ・ For simplicity, we shall assume that the customer’s arr輌val distribution fbllows Poisson distribution with mean arrival rate R. Now, put ρs=λ[M.P.T.]. It is clear by the preceding results that if the system is ergodic, then ρs<1ON AN. EVALUATION OF CHARACTERISTIC BEHAVIOR CONCERNED
27 holds. Thus we shall show ollly that if the relation ρs<1 holds, then the§ystem is ergodic. In order to prove this statement, we wi皿use the fbllowing Foster’s Theo]rem[2], Lemma(Foste〆ぷTheore〃1) 」Let (Pij) (i,ノ=0, 1,2, … ) be the in17nite st・吻ぷがc砲〃Zxσ’んwεm.メ〃ぴe吻〃a∬醐θ’hat the s[ stem is irredticible and aperiodic. The syste〃τis ergodic if there exiぷtぷanon−〃egati昭solution of the inequalitieぷ. ΣPiiy;一≦γi−1 (iキO)
ゴ=0 ぷueh that ΣP。ゴyj<◎。. ト Then we have「the fbllowillg Theorem. [Theore〃T] The system is ergodicぴ ρs<1. [PROOF] x(’)………the size of first queue at the instant’. (include a customer be in first station) y(t)……… the size of s㏄ond queue at the instant’. (inc1Ude a customer be in s㏄ond station) Letτ1,τ2,… ,τn,…うdenote the.instants of the su㏄essive departures from the 丘rst station. Define ’ x。−x(τ。+0) み=y(τ。−0). We assulne the Poisson arrival and the.exponential servi◎e distributions. By the symbo1(1,〃1)we denote the state the size of the五rst queue………(1−1) the size of the s㏄ond queue… …(〃2−i). The process z。・==(Xns Yn) fbrms an aperiodic and irreducible Markov Chain. So, we consider the transit輌qn probabiHties to all possible states from tlle states (1+1,0),(1+1,1),…,(1+i,M+1). That is,28
T.MAKINO
(1+1,0) (1+1,1) (1+1,2) .●● (1+1,i) ⋮ (1+1,M) ︵O。、︶ (〔 g、︶00
丘10
ゐ10
丘褐
え司褐 ︵N。∼︶ ︵否、∼︶ ・ . ・・1・L
1た00
000
丘10
た1
i
(一 {ミ.じ 0 0 0 . ’ kt+11 k8 k3’11 kS−2i ,瑠+’1{[
嚇一1ピー・
H
L
l ]、 F l l I 0 ︵O.一ま︶11
瓦21
丘 1k? k{+1 ﹁ (】 イま︶10
ゐ ︵N.︼+、︶ 001
瓦11
此 (否 ミ一ま︶21
左 ●81 丘01
瓦11
瓦 (一 {ミ、[ま︶ 1 1 ︵O、、+、︶ (〔 D、ま︶ 、 } 「 0 0 k{“i 一2 ・81 た 0 ‖U 0 ー=ー
0i一
0︸
ー
1−1ー..ー1駆9
1渥・}
1
2.豆 丘 3◆フ 丘た81iた㍗・同ザ十
ロ
ー2゚
1:m堰c k;一一1 ]i’ ︵N.、ま︶ (ひ D、ま︶ 0 0 0.ノ 丘 0 . . .1
−’﹂ ん 1 一 ・3.﹂ 〃 ,‘・7’ 瓦 0.’ た 2 一 ・3.ブ ん (一 ¥ミ.、十∼︶ 0 0L
0 0 liザ+1}ザ1ザ一ll i i ザー21 1一 た 0.﹂ Whereん∼is de丘ned as the conditional probability that the number of arrivals to the first queue is/and the number of departures from the second stage is匡between the time of the dCpai加re of a customer and the time of the next customer departure 丘om the五rst stage, co皿ditioning the states be in(1十1,0),(1十1,1),…,(1十1,ハの at the time of the departure of the fbrmer customer from丘rst stage. Pr㏄isely they must be written as kノ(1十1,0),kji(1十1, 1),… , kノ(1十1,ルの. However, we shall omit the symbols in( ), since confusion will llot occur. Let P{1+1,ml},σ+∫.御る} denote the transition probability to the state(1十ノ,〃!2).from the state(1十1,〃1). Then the symbol in the table above means 、. Pσ.・.m、),α。∫.m,)−kim・ト1−m・(1+1,〃1・) (f()r m・+1≧〃1・) −0 ’(for〃1、+1<〃t,). Now we put び”=(M十1)x鈴十ア%. Thell the process{vn}is equivalent to the process . ・ Zn=(Xn, Yn). Therefbre, the{(M十1)(1十1)十i}−th row of the transition matrix is shown as fbllows. State … i(M+1) (M+1)2i×1+ぴ×1+1
‘ (ハイ十1)(1十1)十i[
1 koi+1 kol (M+1)L_.』 ×∼+2P1
koi−1i嬬R。畑!、
l l
⋮ 1 kji+1 kji (M+1) ×(1+ゾ)+2 kji−1 ● ・ ・ ● ・ ・ON AN EVALUATION OF CHARACTERISTIC BEHAVIOR CONCERNED
29 Let ρs≡λ/μs, using the value ofμ5. Whereμs is a r㏄ipr㏄al of[M.P.T.], i.e. 1 μs=[M.P.T.]’ The distribution function of the passage t㎞e through the丘rst stage conditioned the state of the system be in{(M十1)(1十1)十匡}is delloted by珂[(脇1)ぱ+1}÷ξ}(x).(The subscript{(M十1)(1.十1)十i}will be omitted later.) − The probability」Pn that there wiU be precisely n new arrivals during a single occupation time fbr the丘rst stage is.・ Pn−s:e−…(λ詔”跡ω・
And ヨ〃・⇔ヨ{・・1:e−…(λxln!)1・F(・)ト・・∼:・・在ω≦… Letting 〃・・・…一・一・・+・一…−u−・一・r三ρs(・≧・)・ we can see that the condition of the Lemma(Foster’s Theo醐holds. B㏄ause the inequalities in regard to{(M十1)(∫十1)十匡}・th row, co Σρ《∫μ」=ク〔M.1}“.・).‘,(M.・)1・U(M.・)汁ρ(M・・}σ・1}・《,(M・1)ハ・1’U(M+・)鍵・・+’・’ ∫=0 「』,、・[1・{k♂・・+k・・+…+k・・)+(・・‘+1+k・‘+…+k・°)+…} +墓ノ・(・ノ・・撒…+k・・)] ≦1−,、(’+・・)イ圭;、−1−・・…)…−1・and
eo Σρ。∫〃」<。◎ ∫=O holds. This complete the proof. Thus we can derive the ergodic condition fbr the two stage system from the value of Mean Passage Time. In the case of no waitillg room, we n㏄ed皿ot to take into consideration that the service distributions at two stages are exponentia1. However, in the case where the a皿owable maximum size of the waiting room is equal to or greateぱhan one, we assumed exponential service distributions. This assumption has an advantage when we observe at the rellewal time instant, the probability of finishing the second stage service is independent of the past history of.the system. The results obtained are easily extended to the three stage service system(ass㎜一 ing exponential service).30 ‘ T.MAKINO
The procedure of analysis fbr the three stage system goes on as fbllows. . ・・p・⇔P鑑一睡1一麺Llミ翻ト…聯…−i=…⇒…p・・
queue (size:・M−D (size:∧r−1) Fig. 4 Similarly to the two stage system, we use the notations x(’)…・…・・the size of the丘rst queue ’ (include a customer being served) y(t)……… the size of the second queue (include a customer being served) z(’)……… the size of the third queue (include a customer being served) and de丘ne 』 ・ Xn−x(τ。+0),γ。一γ(τ仁0), z。−Z(τ仁0). Considering the random variable Vn−(M+1)(N+i)x。+(N+1)Y。+z。, this process is equivalent to the process(xn, Yn, z.). And in this case we. putthlg . n U‘M’1}(脚力=U(酬N+1}カ+1==’”=U・M・・){N・…+M⑭「一ρs instead of U(”+1’”=U』’・+1=1”=u・M・…+M=了二万:・ ’ we call conclude that the necessary and suMcient condition that the system be ergodic is . ρs<1, since {ui} is a non−negative solution of co co .Σρi∫物≦Ui−1 (f()r i>0), ΣP。iuゴ<。。. FO j=O In general, we can obtain by the similar argument, the ergodic condition fo’r the mUlti−Stage SyStem iS . ρs<1. That is, the ergodic condition of a system is obtained easily using the value of [M.P.T.].』 Jn conclusion, the author wishes express his gratitude to Prof. Y. Tsumura, Asst. Prof. K. Hayashi,. Dr. H. Hatori and Mr. E. Suzuki who have given valuable advices and suggestions to h㎞. Appendix. The Table 4 shows numerical values of(1−」PB), that is, the maxh皿um utilization factorρm。。 for various numbers ofルf(0∼3),1V(0∼3)andμ‘(’=1,2,3). Now let us note that 、・’:’ON AN EVALUATION OF CHARACTERISTIC BEHAVIOR CONCERNED
3r. [M…i・コ「寺・]、 holds. (Where PB denotes the mean blocking.rate at the丘rst stage.) The fbllowing descriptions ca皿be derived from the Table 4; In the case where the order of service stations are interchangeable, we can recognlze that the maximum Value Of pmax iS obtained by aSSigning the ServiCe、which has max−・i㎜ms瓠vice rateμξ@=1,2,3)to theぽond stage.
This statement in a㏄ordance with the description in[21エ The calculatiOns fbr the Table 4 was cut in haif, b㏄ause we assumed that the, reversibility of[M.P.T.]is satis丘ed. The fb皿owing is an example. The Mean Passage Time and the mean bl㏄king rate in Figure 1 are described as [M.P.T.(1)コ・・d P.(1)・esp㏄ti・ely,・nd i・F迦e 2…[M・P・T・(2)コ・・d P・(2)・ Thus we have ・−x(・)一÷・一(1)・ since [M.P.T.(1)]=[M.P.T.(2)] holds.(W…eρ一・ω一1−P・⑦(f…−1・・)…一=‘1)・ ’
[EXAMPLE] In the case of M=O, N=1, XL1 ・= 1, μ2=1.2, μ3=1.4,we have
ρiitax(1)=0.6988, using the Table 4. Therefore, in the case of M=1, N=O, μ1=1.4, μ2=1.2, μ3=1,we have
・一(・)r!、…6988・・一・・4991・ In thiS way, if We knOw the value Ofρmax in the caSe Of M・−i,N一ノ, then’翌?@can easily obtain the value ofρ皿ax in the case of ルt=ノ,N= i.T.MAKINO
32
OOON卜箇●N寸一②ooい一〇
ひひo②嶋o●寸←吟oの寸o
ぬooひ②N←oひ崎←Ao−o
th寸嫡②←r∼oo◎o oo o◎ooひ・o
・ ・ ・ ・ ・ ・ ・ ・ … .ひ ・
一
8oo一寸ooo一寸い②②ひ一N
←めNぬ一め←[寸一ぬトトo
ぬoo ooい一め◎〇一否め②←o◎←
否寸い②卜←←◎o.oo oo oo oo.
・ ・ … . ・ ・oo . ・ . ・ひ
゜。 DNいo吟NNめ寸[〔ooo■ひN
②寸othooo∩ひo←ON一寸
い◎o◎oめoぬ◎oo否寸②←oo②
th寸い②卜←←oo oo・◎o・◎oひ
・ ・ ・ ・ … . .o◎ . oo . .
②.No◎o寸ひ、◎画②一吟一ゆひ否ぬ
ONooo寸②◎o寸寸一寸いぬ②
吟co←めo寸卜ON寸嶋②トめ
否寸い②卜←卜oo◎o oo oo◎◎ooひ
寸.N否一←めσ、oo寸寸めぬ←卜◎oぬ
い一ぬ②ひON←←否②ト②②
いoo←寸ひ寸←ひ一■寸い②マ
th寸い②②卜卜←oo◎o◎o◎o ooひ
N.N否ひ一いトぬn②ひい否ひ。。否
寸OO N一め[寸oo←[“)②ぬ箇
のト←寸ひ否◎OOONm⑰い∩
[寸頃ゆ②卜←←oo oo o◎oo ooひ
O.Nooo−oo寸ooいoo[否いo一〇〇
のぬ←寸ぬ否め②いONPt一い
頃卜②而OONめトA−N面寸一
否・冶ゆ②←卜卜卜oo◎o o◎ooひ
・「轣@・ 呼 。 . . . . . . . . .
oo D一一ぬN寸ひぬト←N的◎o寸■o
一一〇い面Ooooo◎N画否一N
い←MりN←−m②←⑳o−N⑳
th寸1ハ●②卜←卜←卜OO◎O◎O OO
●.一否∩∩Ptooo一い℃②一ひNN
ooめON←Noooo寸卜o◎●寸ひ
寸②ぬ一いひ一否い②卜ooひい
め寸ぬゆ②ゆト←卜←ト←←oo
寸.︼oo◎oひぬ「守小o[oo−OON←
寸㍉◎めN寸いoo←−NPtひ◎M
寸い否ひ否②OOONひ寸寸嶋一
の寸冶い②、◎ゆ←←←←←←oo
N.一一寸②一N←②o◎←[一寸←o
トー否寸o●●−th否一トNO
∩寸一②ON寸∼◎卜oo◎oひo嶋
否寸めぬ②N◎②②②②りwo←←
O.︼の←◎。o卜寸◎・NN[一ト②N
めトoooooひ寸②めNO◎N②一
N〔←N寸②ooひo〔−NN●
th寸寸いぬぬ婚1の②v②②N◎②
゜。 DOP;繰R工8#零雲$雲寺R8苫
oトPtいトo◎ひoo−一一Pt寸
否[寸寸寸寸寸め嶋ぬ吟いい嶋
●.O←否oいトひ箇寸oo卜一th柏N
o◎一い寸oいOOON寸い●寸
②o否寸いw②②卜卜←←ト。。
Nひ市否thの画否画否面[否 ・
・・. . . ・ ・ . ・ . ・ 守 ◆ ・∩
寸.O 8 °o DN ②.N 寸.N N.N O.NON’守N◎o◎
▼−4、一⑯ ▼−4 、一叫 一 ゜o DO ②.O 寸.O ︵ぷ←@
i報
[釣㌔.ポー−£ぷ§・ほ︶ H、、日⇔﹄o百鳴角=o=尉N一一=P巨5日⋮×儒芝 .⑰o呈吋8 、 O”≧.01ーミ ノマON AN EVALUATION OF CHARACTERISTIC BEHAVIOR CONCERNED
33
’.−t
ひ寸柏N一〇〇卜■●寸ぽひ寸ひ
◎oひひぬひ←NONOぬトトひ
鴇等零8fl繰8窃おs認訟まま
留誤8x苺8雲88s自訳菖葛
累寺繋$91F:89給溜器台8ま
畠まお?工$象F留$s鴇鴇自
恰等訳81!苦舎9葛お認9まま
ゆトひひoo②い柏②ひo◎市否
oo oo←Nひひト寸冶N←oo◎o●
いoo oo●一、◎ひN寸、⇔←ooひト
柏寸い●卜←←OO OO OO OO OO ODひ
●NOON崎②oooひ②ひ否卜
oooo●一トo寸N一トーNN←
め◎ooo、◎一、⇔ひN寸めトcoひ②
否寸ぬ●←トト◎oco◎oooooooひ
[寸Nトo一②o一否寸e∩否否
co卜■∩oo寸②◎o寸⑰o吟寸[い
めo◎ooぬ一嶋oo一箇い②卜ooめ
市寸嶋②←卜←coo◎◎oooo◎ooひ
舎⑦潟等9台8#需鵠99ま$
いooooぬo「守OOON否い②②面
否寸ぬ●トトトoo oo oo co oo◎oひ
一ひ一いoo−N卜NNめト柏●
卜否oo←oo←②◎o②o〔ooo oo
いoo←寸ひ面◎◎ooNth寸寸o
th守め②●←←←OO OD oo oo ooひ
oo←一否②寸寸Nい●[N②v
擦§§§§E蓑§9$§§菖9
o②一■NいひON←N面柏o
柏一吟否寸th寸一否NO●−o
め←崎一ぬOOONひ寸いいゆo
画寸め●②ゆ←←卜卜卜←←oo
ひひ一←ひ否寸いトooひ一ぬ
寺菖鴇1!8☆等s$$F苺巽9
th寸ぬ》、噸②②②●●●ゆ②卜
ひぬoo冶oo卜卜箇寸めoo←N否
一〇〇〇〇〇〇▽◎o②N●⑳一[めwo
需:等等;窮需蕊蕊蕊鴇$恰$
吟〔い’トひ←oトOD●N←一◎o
宗等$鐸9鵠8s留台まま宗ま
Nひ面柏否画否[的[門ひ否吟
8 ◎o DN ②.N 寸.N N.N O.N ook
②一 寸一 N一 〇.一 o。 DO ②.O 寸.O 8 ◎o ・N
② ・N
寸 ・N
N
・N
o
・N
’ o◎ . 一 ゆ ・ 一 寸 ・ 一N
. 一o
. 一 ◎o ・o
●.o
寸 ・o
腸 ト 鯛 卜 一11之.01ーミT.MAKINO
34
,ほt§亀§‖§菖Sξ塁§ま§
一
◎oN
…購顯§§詩雛§§§薬
⊇購§§馳§藁詩竃詩舗員
ひトoo◎N一ひひゆooトマぬトひ
ひ②◎ぬ寸⑳②ひ◎o寸o◎◎◎トー
ODひひ噴画←N寸一ひ否無吟め寸
熔等訳89担89お£認$象ま
O.Nひr守一、ooo−o否■の崎o◎◎o
鱈購§§§§設ss§ま
◎o D一oo−NO吟いト●一■nひooooo
ooひo◎否ひNめ一Noo−N−Pt
いoo oo●一②ひN寸めトooσ、◎
否寸ぬ②卜←卜oo oo OD◎o oo ooひ
②.一←否嶋←めOO[②トOOOONト
◎ooo②ひ寸ぬトーoめ◎o oo◎oo
いOD ooぬ一いoo一柏寸■∩▽←否
⑰寸い②卜←卜oo oo oo oo oo ooひ
寸.一雷$R自守雷8霜ま吉☆台㌍ま
ぬ◎ooo嶋o寸トひON[寸マ1ひ
否’守1の●トト←トoo oo oo oo oo oo
N.一否卜寸寸寸寸②oo卜②マe卜否
←NめトN寸卜寸トトい〔②否
めco卜面oo−[嶋●卜oo a⑳柏
th寸め◎◎卜←トトトトトトco
め一◎、一◎ひN−o⑳寸oooひ
菖謬ぽ白鴇EXに葛ぶs白88雪
の寸嶋②◎②②㊨②②卜卜トト
o嶋②②←一トーN吟一寸寸寸
い一めめN否o、寸←ひ一N画o
⑰寸⑳N寸吟い◎②②卜←卜゜°
市寸寸ぬいいいめぬい吟めいい
O.︹ ゜o DO 達.O 8 °o.N ②.N 寸.N N.N O.N oo ②.一 寸一 N一 〇一 ゜o.O ②.O 寸.O d11≧^01ーミON AN EVALUATION OF CHARACTERISTIC BEHAVIOR CONCERNED
35i8ま白$$8x等FR8鴇:8
8累寺$$窪:888S$まま8
一loOON②N②〔柏ひ寸ぬ〔卜←
◎。i$訟ま8沿繰8蕊$繰ま白恰ま
べi『’“t口9hh°『°『’°『°『°『9〔『
寒1購蕪§§§舗舗§ま§
looo〔②一寸ひo②ひoめoひ
寸i$$ま$☆R8蕊$苺ま白雲ま
dln『口9t‘:亡竺竺『QO今今〔
o◎o〔いON②oo一⑰寸00NN
ひひo②ぬo吟面②寸σ、−Nぬ
累等$$9忘8雷給s器まま象
N.Nooo一寸coひ一〇箇寸[●o◎一
$台ま$蕊$8需$fr2認8田象
否寸め②←←oo oo oo oo ooひaひ
O.NOトひ一のN一卜卜嶋一一一ト
⑳ひひ●寸ひ寸一ひ一②ooo9ψ
めoo oo②N②oひいトooひooo
否寸の②卜卜◎o◎ooo◎◎◎oooひひ
゜◎.一o②卜②否②o、a寸②●No◎oo
σ、ひひ1ハ画卜一〇◎o←一否N●
熔寺訳$㌍担89協溜留台まat
②.一ひ寸o−ooひ〔o◎−N否ひトー
ooひひ寸oひ←NひひNN−oo
いoo oo②NゆひN寸ぬトooひ吟
th寸ぬ②←←卜oo oo◎o oo oo ooひ
寸.︼ooひN寸卜一②oo’
ゥoo一、r嶋②
OO OD←o寸崎いoo②ON一〇〇N
めoo OD②一めo◎OPt寸ぬ●②N
否寸めo←卜←OO OD oo oo oo ooひ
N.︼めoo一冶o一寸一ゆ◎o画ぬ一
留お9自ま需SIRま88巴RC
n寸ぬ、◎●←卜←←OD◎o oo oo oo
O.︹ぬoNい一否ひト寸一寸卜寸ひ
トoのoゆひ・守ぬ否ひひ②ひ②
寸oo②N嶋←ひo−−NNN寸
否寸ぬ②②②②卜←トト←←ト
◎o DO◎②ooひo嶋o[ON−◎oめ
一い面否o◎oo否卜o、−Nひnoo
ぬぬ一寸めゆトトト◎o◎ooooo◎o
否寸めいぬいmぬい嶋いい硲硲
ゆ.O葛81}x$田謹9完2葛奏器ま
N卜ooひひひひひζh⑳ひζ“ひひ
面[否の[柏否柏否mの的否柏
寸.O 8 00 DN ②.N 寸.N N.N O.N ゜〇 ②一 寸一 N一 〇.一 ゜o DO ②.O 寸.O 吟H之.01ーミ ・T.MAKINO
36一
8ひい寸[∩N②一寸寸⑰oo否
[ひひ←oぬ←−oo〔一ひの②
oo∩ぬ寸〔いoo−N寸いめ②ひ
否崎o←oo◎o ooひひひひ⑳ひひ
゜o DNOON◎oいNO一のOOoo噂[Noo
ひ⑳◎owoひ寸②ひ②ひひト⑰寸
oo[め寸oめooON[寸め▽ひ
thい②←oo oo ooひひひひひひひ
②.N卜◎ooひ②一一thい寸Noo卜②
画ooooぬ←Nマト寸←卜寸ON
◎o否め寸oめ◎OON⑰寸い◎ひ
.め②←oo oo◎oひひひ⑳ひひひ
寸.Nめ一〇〇い寸ぬ〔一一ひ②一〇否
酌◎o▽否寸ひ〔寸一否否一←ひ
oo否崎、守oマOO O N否寸いめoo
箇い②←◎ooooo⑳ひひひひひひ
N.NNN一ひoいo情o②N②マー
否←硲ONいN◎ひ、◎oo◎oめ一寸
oo[め寸o寸トひ一N[寸ぬoo
面崎◎←oo oo oo◎oひひひひo、ひ
O.Noo◎o寸oo◎oめ◎◎ひNいo◎−oo oo
NめNゆ②ひひ一〇〇〇ひトNめ
◎o柏恥否ひひ②ひONN否寸卜
.めw←←OO OO oo⑳ひひひひひ
゜o D[自熔8さ窃8台δ訳旨$鴇宗R
oo[寸のoo Nいco⑳〇−NN②
のい、◎卜←OO oo oo◎oひひ⑳ひひ
②.一◎o←oo oo o卜N一卜卜一②oo←
oひoひめ否一一め②ぬ〔●◎o
OON寸一←一寸ゆ←ooひoo■
箇ぬ②←←◎Ooooooooo◎oひひひ
寸.︼、◎一寸oo画否ひのめN②寸ひth
ooひ◎〇−N←一ひNNひいひひ
富Sl 9繰鴇廷完9蕊鵠奏鵠鵠鉛
N.︼寸◎o否吟寸め←o寸N一ト∩市
⑰o②o[NNトトめ一いひめ
Pl言$$1:lf!旨に廷1!1888訟
O.一卜OcoいNoo卜←ひ◎ooo一ひ
い←いい←OO N−oo∩←o⑰oo
②oo②〔寸②◎o⑳ひoo−一N
th寸い∨)②②ve②②←卜←卜←
゜o DO一←ゆ◎寸寸崎箇←−oマ寸寸
ゆひNNO−oo否▽ひ一Ntho
寸〔戊⑳Nマめい、◎②②←←卜oo
←り寸寸いめぬめぬ1のいめ1nいぬ
②.O1寸Pt否o←o一寸否⑳寸卜oい
寸ま等繰完8ま象まままま88ま
o.N[門ひ■めひ[の否Cfi否否∩
1 ● ・ ・ . ■ ● ■ ■ ● ● . . . 右 8 00.N ②.N 寸.N N.N O.N ゜。 ●.一 寸.一 N.︼ O.︼ oo DO ▽.O、 寸.Oシβ
一‖﹀∼ 、一HミON AN EVALUATION OF CHA[RACTERISTIC BEHAVIOR CONCERNED
37 8N寸NO←NOoo否ひいNNO
寸〇−o否ひNい否②ゆ寸oo
oo寸②め〔ぬひ一[寸め、◎←o
否い、◎←oo oo ooひひひひ◎sひo
一
|N寸oゆ一寸oトーN−←卜卜
『葛?お等se留宗苫9等yl$田ま
N∩頃▽←OO OD oo⑳ぴひ⑳ひひひ
ジ§§§§塁§§三旨奪ま§§§
NlOり⑰亡竺曽竺〔『〔○〔〔9
∋§§§§§§嵩1§導苔旨§象
N∩い②卜◎ocoooひひひひひひひ
PtiO竺9ビ゜『竺゜『99『99一込〔
sl§§§§9蕩嵩§§§旨§日§
i一ひひ②N←●←②否ひ画一一
oめ⑳oσ、oo−←ひ②ONひ←
寸ひ◎o②トー[いN寸寸一℃②
co thめ寸oぬOO O Nの寸ぬめoo
thい②卜co oo◎o⑳ひひひひひひ
o。 D一Di㊤留99$$$蕊等$9Fls鵠
;4’ llll『旨さ完;li Sii$ii{1)1零穗igi li
1い否oo一寸oぬcoゆ▽oNN一
寸i2桧訳18鴇葛葛象9曽☆自訳
一1否い②卜←◎o◎ooocoひAひひひ
!OO N②ひo“)ひoo②一N②oo②
Nl9蕊專旨P沿写零1:完器ま8寒
口i〔り『亡卜゜°°°。。。。。。。。。。◎。⑳
一 一 ͡ r 1 . 一:一一〇のN◎寸寸〔oo⑳一一N−
『;完ll:8等iQ 8台巴☆冨鴇留$
一’吟い《)←卜卜←◎OOOODoo◎o◎ooo
いo◎⑳N←一吟←寸寸否ぬNひ
竺i{188等決㊤:9R自自茗況●
⇔ひめめゆ②②卜←卜←卜卜←卜
1 右
. . . ・ ■ ・ . ■ ■ ■ , ・ ・91§§§奪§9邑{il 2§§菖§る
O・9 「r □ ℃ □ □ Lt’} □ 竺 〔 vr 竺 vr □
’ゆひ、◎②トひN卜ON寸め②o
〔は8葛まままままままままま象
Ol『竺○『e”1『『○『『で○○『
三二±ニニニ災ささ竺8
゜o.O ②.O 寸.O / 、//﹁・↑ 。,g,/
N‖≧ばー﹂ミT.MAKINO
389葛§8PlぶR留需沿沿#88
00寸②ぬ一いひ一門寸い②卜o
柏ぬ②卜oo◎o ooひひひひひひo
一
8N寸Nひ②一←いo−一◎o◎oい
寸〇一ひ⑰ひ一いひoゆ否ひひ
oo寸ゆ寸一ぬ⑳一否寸め②②ひ
柏ぬ②←oo◎o ooひひひひひひひ
oo DNN寸Nひぬoゆ而◎oひひいいN
寸o一ひ否ひ一ぬNいの市ひひ
◎o寸②寸一いひ一否寸い▽②ひ
柏い②←oo co ooひひひひ⑳ひひ
ゆ.NN寸N◎o寸oo寸oぬめ嶋一一卜
寸〇一ひth oo一いNいぬ∩ひ◎o
oo寸②寸一ぬ〔一[寸ぬ②ゆひ
市い②←◎oooooひひひひひひ⑳
寸.NN寸一卜一寸ひbひ⑦←enNoo
寸〇一ひ画coo寸一寸寸Noo←
◎o寸②寸一ぬ(あ一[寸ぬ②㊨ぴ
否ひ②←◎o◎o◎oひひひひひひひ
N.NNthひ寸卜◎〇一ぬト②否oo←一
寸ooひN←o面oth否o②②
co寸②寸一ぬひ一th寸ぬ②②ひ
否め②←oo oo ooひひひひひひひ
O.NNN②OD卜寸否寸寸o崎ひ②←
寸oooo一②oo−co−o←∩N
oo寸②寸一いoo−N寸めぬ、◎ひ
柏ぬ②卜◎o.◎oooひひひ⑳ひひひ
゜◎.一