174

CHAPTER 6

ANALYSIS OF A BULK QUEUEING SYSTEM WITH

CONTROLLED ARRIVAL AND MULTIPLE VACATIONS: A
SIMULATION APPROACH

6.1 INTRODUCTION

Nowadays, “lean” systems are much preferred by the organizations

to simplify their work. Many of the organizations are doing research and
improving their operations using simple analytical tools instead of involving
people at all levels. Simulation is an applied technology which is mainly used
for analyzing and solving problems and it is a most popular approach to find
the performance measures of complex queueing models. Simulation models
work appropriately because of the analytic power of queueing theory which
underlies and enables them. Applying simulation begins by being clear on the
definition of the problem, the reasons for simulating, and the expected
outcomes. A Non-Markovian bulk queueing system with state dependent
arrivals and multiple vacations has been studied by Ramaswami and
Jeyakumar (2014). They used ARENA software to model the system and
derived some of the performance measures. Recently,
Moazzami et al. (2013) have focussed on modeling the behaviour of a petrol
station and they have used WITNESS 2004 simulation software to model and
analyze it. In the literature, only very few authors have given attention on
analyzing queueing models using simulation. The primary focus of this
chapter will be on modelling and analysis of the proposed queueing model
using simulation technique.
 175

In this chapter, M X / G(a, b) /1 system with limited number of

admissions and multiple vacations is analyzed. The service is done in bulk
with a minimum of „a‟ customers and a maximum of „b‟ customers. The
operator provides batch service when the queue length is greater than or equal
to „a‟ and less than or equal to „b‟. If the queue length is greater than „b‟, then
„b‟ customers are considered for service and the rest are kept in queue. On
the other hand, when the number of customers is inadequate to process i.e.,
less than „a‟, then the operator goes for secondary job (vacation) during its
idle period. On completion of a single vacation, if the queue length is still
less than „a‟, the operator goes for another vacation (multiple vacations) and
so on, until it finds „a‟, customers in the queue. But all the arrivals are not
considered for service all the time. During the service period, the arrivals are
accepted with a probability α whereas during the vacation period, the arrivals
are accepted with a probability β. For the proposed model, simulation is
performed.

The application mentioned in chapter 2 involving the process of

cone winding in a textile industry, is a suitable example for the proposed
queueing model. Autoconer (server), the machine used for winding process
(bulk service), is operated to the maximum capacity of 60 ring cops
(component/customer) called its upper limit (b). The machine will be
operated only if sufficient amount (minimum capacity „a‟) of ring cops is
available in the queue to start the winding process; otherwise, the operating
cost increases. Once the process is started, the bulk operation has to continue
successively for many batches.

On completion of the winding process for that batch, the process

continues if the number of ring cops available in the queue reaches atleast the
minimum value „a‟; otherwise, the operator goes for vacation. The yarn
obtained from the winding process is taken for packing. After packing,
 176

further processes like warping, sizing, pirnwinding and weaving which can be
considered as secondary job (vacation) are carried out by the operator. Upon
completion of secondary job, if the number of ring cops available in the queue
reaches atleast the minimum value „a‟, the operator starts service; otherwise,
the operator goes for multiple vacations.

In order to deliver the processed components in time, the operator

may control the new arrival (limited number of admissions) to some
probability. The operator accepts only α percent of the arriving components
when the server is under the winding process (busy), β percent of the arriving
components when the server is under the process of packing, warping, sizing
etc., (vacation). The above situation existing in the textile industry can be
modelled as M X / G(a, b) /1 queueing system with limited number of
admissions and multiple vacations.

In this chapter, the simulation model of the proposed queueing

system has been developed to analyze the performance of waiting lines at
different resources in a production system. The core purpose of this chapter is
to draw attention to analyze the system performance metrics such as
throughput, average components, average waiting time, overall production
time, and resource utilization involved in a manufacturing industry. An
efficient simulation model has been constructed and analyzed using
Flexsim 2017, powerful software for simulation and justified through
numerical illustration.

6.2 MATHEMATICAL MODEL

The probability generating function of the number of components

in the queue at an arbitrary time epoch and expected queue length are
obtained in this chapter. The notations used are as follows:
 177

„X‟ - group size random variable of the arrival

„λ‟ - Poisson arrival rate

„gk‟ - probability that „k‟ components arrive in a batch

X(z) - Probability Generating Function (PGF).

α, β - probabilities that the arrivals are accepted during busy period and
vacation period respectively.

S(x) (s(x)) {S ( )} [S0(x)] - cumulative distribution function (probability

density function) {Laplace-Stieltjes transform} [remaining service
time] of service.

V(x) (v(x)) {V ( )} [V0(x)] - cumulative distribution function (probability

density function) {Laplace-Stieltjes transform} [remaining
vacation time] of vacation.

Nq(t) - the number of components waiting for service at time t and

Ns(t) - the number of components under the service at time t.

The different states of the server at time„t‟ are defined as follows:

C(t) = (0)[1] if the server is (busy)[on vacation]

Z(t) = j, if the server is on jth vacation starting from the idle period.

To obtain the system equations, the following state probabilities are defined:

Pi,j (x,t)dt = Pr{ Ns (t) = i, Nq (t) = j, x ≤ S 0 (t) ≤ x+dt, C(t) = 0 },

a ≤ i ≤ b, j ≥ 0 (6.1)

Q j,n (x,t)dt = Pr{ Nq (t) = n, x ≤ V 0 (t) ≤ x+dt, C(t) = 1, Z(t) = j },

j ≥1, n ≥ 0 (6.2)
 178

6.3 STEADY STATE QUEUE SIZE DISTRIBUTION

The steady state queue size equations obtained by using

supplementary variable technique are as follows:

b
−
d
P (x) = − λ Pi,0 (x) + λ (1-α) Pi,0 (x) +
dx i,0
P (0)
m=a m ,i
s(x)

∞
+ 𝑙=1
Q 𝑙,i (0) s(x) ; a ≤ i ≤ b (6.3)

j
−
d
P (x) = −λ Pi,j (x) + λ (1-α) Pi,j (x) + α
dx i,j
Pi,j−k (x)λg k ;
k=1

a ≤ i ≤ b-1, j ≥ 1 (6.4)

j
−
d
P (x) = −λ Pb,j (x) + λ (1-α) Pb,j (x) + α
dx b,j
Pb,j−k (x)λg k
k=1

b ∞
+ Pm ,b+j (0) s(x) + Q 𝑙,b+j (0) s(x) ; j ≥ 1 (6.5)
m =a 𝑙=1

b
−
d
dx
Q1,0 (x) = −λQ1,0 (x) + λ (1-β) Q1,0 (x) + P (0)
m=a m ,0
v(x) (6.6)

b
−
d
dx
Q1,n (x) = −λQ1,n (x) + λ (1-β)Q1,n (x) + P (0)
m=a m ,n
v(x)

n
+β k=1
Q1,n−k (x) λg k ; 1≤ n ≤ a-1 (6.7)

−
d
dx
Q1,n (x) = −λ Q1,n (x) + λ (1-β) Q1,n (x)

n
+β k=1
Q1,n−k (x) λg k ; n ≥ a (6.8)

−
d
dx
Q j,0 (x) = −λ Q j,0 (x) + λ (1-β) Q j,0 (x) + Q j−1,0 (0) v(x) ; j ≥ 2 (6.9)

n
−
d
dx
Q j,n (x) = −λQ j,n (x) + λ (1-β) Q j,n (x) + β Q j,n−k (x) λg k
k=1

+ Q j−1,n (0) v(x) ; j ≥ 2, 1≤ n ≤ a-1 (6.10)

 179

−
d
dx
Q j,n (x) = −λ Q j,n (x) + λ (1-β) Q j,n (x)

n
+β Q j,n−k (x) λg k ; n ≥ a, j ≥ 2 (6.11)
k=1

6.4 PROBABILITY GENERATING FUNCTION OF THE

QUEUE SIZE

The following Probability Generating Functions are defined in

order to obtain the Probability Generating Function (PGF) of the queue size at
an arbitrary time:

∞ ∞
Pi (z, θ) = j=0 Pi,j (θ) zj and Pi (z, 0) = j=0 Pi,j (0) zj , a ≤ i ≤ b

∞ ∞
Q j (z, θ) = n=0 Q j,n (θ) zn and Q j (z, 0) = n=0 Q j,n (0) zn , j ≥1 (6.12)

∞ ∞
where Pi,j (θ) = 0
e−θx Pi,j x dx and Q j,n (θ) = 0
e−θx Q j,n x dx are
Laplace - stieltjes transforms of Pi,j x and Q j,n x respectively.

Let P(z) be the probability generating function of the number of

components in the queue at an arbitrary time epoch of the proposed model.

b−1 ∞
Then, P(z) = i=a
Pi (z,0) + Pb (z,0) + 𝑙=1
Q 𝑙 (z,0). The
probability generating function is obtained as,

β bi=a
−1
c i z b −z i S α λ−λ X z −1 +
β S α λ−λ X z −1 +
V (β(λ−λ X(z)))−1 a −1 (c z n )
n =0 n
α z b −S α λ−λ X z
P(z) = (6.13)
αβ (λ X(z)−λ) z b − S (α (λ−λ X(z)))

b ∞
where m =a Pm ,i (0) = pi , l=1 Q l,i (0) = qi , and pi + qi = ci
 180

6.4.1 Steady State Condition

The probability generating function has to satisfy P(1)=1.

Applying L‟Hospital‟s rule in (6.13), then the condition ρ < 1 needs to be
satisfied for the existence of steady state for the model under consideration
where ρ is αλE(X) E(S)/b.

6.4.2 Computational Aspects of Unknown Probabilities

Equation (6.13) gives the PGF P(z) of the number of customers in

the queue at an arbitrary time epoch, which involves „b‟ unknown
probabilities namely, c0 , c1 , c2 , … , cb−1 . By Rouche‟s theorem, the expression
z b − S(α (λ − λ X(z))) has b-1 zeros inside and one on the unit circle
z =1. Since P(z) is analytic within and on the unit circle, the numerator of
(6.13) must vanish at these points, which gives „b‟ equations and „b‟
unknowns. These equations can be solved by suitable numerical techniques.

The unknown probabilities q0 , q1 , q2 , … , qa−1 involved in P(z) are

expressed in terms of p0 , p1 , p2 , … , pa−1 which are useful to find some of the
performance measures. The constants qn involved in P(z) are expressed in
n ω0
terms of pn as, qn = i=0 bn−i pi ; n=0,1,2,….a-1, where b0 = ,
1−ω 0

ω n + nj=1 ω j b n −j
bn = , n=1,2,…a-1 and 𝜔𝑖 is the probability that „i‟ components
1−ω 0

arrive during a vacation period.

6.5 PERFORMANCE MEASURES

6.5.1 Expected Queue Length

In this section, in order to validate the simulation model, the

expression of expected queue length is derived by using the PGF of queue
size distribution. The expected queue length E(Q) at an arbitrary time epoch
 181

is obtained by differentiating P(z) (equation 6.13) at z=1 and is given by

lim𝑃′ 𝑧 = E(Q)
z→1

b−1 b−1
i=a β ci b b − 1 −i i−1 𝑓1 𝑋, 𝑆 + i=a β ci b − i 𝑓2 𝑋, 𝑆
a−1 a−1
+ n=0 β cn 𝑓3 𝑋, 𝑆, 𝑉 − 𝑓4 𝑋, 𝑆, 𝑉 + n=0 α cn 𝑓5 𝑋, 𝑆, 𝑉 − 𝑓6 𝑋, 𝑆, 𝑉
E Q =
2αβλE X T1 2

where,

S1=αλE(X) E(S), V1=βλE(X)E(V) , T1=b-S1,

S2= αλE(S)X‟‟(1) + α2λ2E(S2)[E(X)]2,

V2=βλE(V)X‟‟(1) + β2λ2E(V2)[E(X)]2,

S3= λ T1 X‟‟(1) + λ E(X) b b − 1 - λE(X)(S2),

f1 X, S = T1 S1 , f2 X, S = T1 S2 -α E(S)S3,

f3 X, S, V = T1 2i V1 S1 + S2 V1 + S1 V2 ,

f4 X, S, V = S3 S1 βE(V),

f5 X, S, V = T1 2i V1 T1 + b b − 1 V1 − V1 S2 + T1 V2 ,

f6 X, S, V = S3 T1 βE(V).

where E(X), E(S), E(S2), E(V), E(V2) are defined in numerical illustration.

6.6 SIMULATION MODELLING

Figure 6.1 depicts the simulation model developed for the

proposed queueing system using Flexsim. The objects used for modelling are
source, queue, exit (sink), combiner and processor. The source in the model
is an object which generates entities (ring cops/components) for the winding
process (service). The inter-arrival time of the components follow
exponential distribution.
 182

Figure 6.1 Simulation modelling of the bulk queueing system with limited
number of admissions and multiple vacations

The arriving components are allowed to join the queue with

probability α, β (limited number of admissions) during the busy period and
secondary job/vacation respectively. Therefore, the components join the
queue with probability „α‟ or leave the system through exit1 with probability
„1-α‟. When the queue length is greater than or equal to the minimum batch
size „a‟, and less than or equal to „b‟, the operator starts the winding process
(in the server/autoconer) for that particular batch of ring cops and continues
the process until the queue length becomes less than „a‟. If the queue length
is greater than „b‟, then, the batch of „b‟ components is taken for the winding
process and the rest are kept in queue. The combiner is used to batch
(processing time is zero minutes) the components before the process. The
flexscript code (in flexsim) used for bulk service is given below:
 183

Flexscript code:

/**Custom Code (in combiner) for bulk service**/

Object current = ownerobject;

Object combiner = current;

Object Queue = current.inObjects[2];

int LowerLimit = current.Lowerlimit; /** Lowerlimit=a**/

int UpperLimit = current.Upperlimit; /** Upperlimit=b**/

if (content(Queue) >=LowerLimit) {

if (content(Queue) >=UpperLimit) {

Table thelist = getvarnode(combiner, "componentlist");

treenode thesum = getvarnode(combiner, "targetcomponentsum");

thesum.value = 0;

for(int index = 1; index <= thelist.numRows; index++) {

thelist[index][1] = UpperLimit;

inc(thesum, UpperLimit);

openinput(current);

else {

Table thelist = getvarnode(combiner, "componentlist");

treenode thesum = getvarnode(combiner, "targetcomponentsum");

thesum.value = 0;
 184

for(int index = 1; index <= thelist.numRows; index++) {

thelist[index][1] = content(Queue);

inc(thesum,content(Queue));

openinput(current);

/* The conditions given on the exit trigger of the combiner*/

Object item = param(1);

Object current = ownerobject;

int port = param(2);

// ************* PickOption Start ************* //

/***popup:CloseOpenPorts*/

/**Close and Open Ports*/

/** \nThe action will be performed if some condition is true (equal to 1).*/

Object involved = current;

int condition = true;

if (condition) {

/** \nAction:
*//***tag:action*//**/closeinput/**list:closeinput~openinput~stopinput~
resumeinput~ closeoutput~openoutput~stopoutput~resumeoutput*/
 185

(involved);

} // ******* PickOption End ******* //

// ************* PickOption Start ************* //

/***popup:SetLabel:hasitem=0*/

/**Set Label*/

Object involved = item;

string labelname = "Customers";

Variant value = getvarnode(current, "componentlist").as(Table)[1][1];

involved.labels.assert(labelname).value = value;

} // ******* PickOption End ******* //

After getting service, the components leave the system through

exit2. On the other hand, when the queue length is less than „a‟, the operator
goes for vacation/secondary job like warping, sizing etc., Source 2 in the
model generates entities for secondary job and after completion of secondary
job, the entities leave the system through exit 3. The flowchart for the model
is given below in Figure 6.2.
 186

Customers waiting in queue (α)

Run
(Service)

Queue length < a

No

Yes

Secondary job
(Vacation)

Queue length < a

No Yes

Figure 6.2 Flowchart for the model

6.7 NUMERICAL ILLUSTRATION

The performance measures are evaluated in order to predict the

behaviour of the queueing system. Across industries and disciplines,
simulation modelling provides valuable solutions by giving clear insights into
complex systems. Simulation is often utilized when conducting experiments
on a real system is impossible or impractical, often because of cost and time.
Since analytical method is a tedious process to obtain different
 187

performance measures, simulation for the proposed queueing model is

performed. The performance metrics such as total number of inputs, outputs,
average content, average waiting time, processing time and idle time, etc.,
which are difficult to find theoretically, are easily obtained using simulation.
To state precisely, a simulation model can capture many more details than an
analytical model such as number of inputs, outputs, etc., in every object/state.
The simulation software, Flexsim 2017, is used for simulation and the results
are obtained using the following assumptions and notations:

Batch size distribution of the arrival is geometric random variable with mean
2 (E(X)=2)

Service time distribution is 2-Erlang random variable with parameter μ

(E(S)= 1/μ, E(S2)=3/2μ2)

Vacation time is exponential random variable with parameter ξ

(E(V)= 1/ξ, E(V2)= 2/ξ2)

Minimum service capacity a

Maximum service capacity b

Probability of arriving batch will be allowed to join the queue during

busy period α

Probability of arriving batch will be allowed to join the queue during

vacation period β

In a textile industry, it is assumed that the process of cone winding

starts with minimum value of 2 units (i.e. a=2) of ring cops (components)
(1unit = 6 ring cops) and the maximum capacity of 10 units (i.e. b=10) of
ring cops (components) (60 ring cops).
 188

The simulated results are noted for run time: 108665.92 minutes or
75.46 days approx. The simulation process has not taken 75.46 days to run.
The operator can fix the running time and increase the running speed through
the run speed option available in the software where it doesn‟t take longer
time to produce the results. In order to obtain the steady state result, the run
time is fixed as 75 days approximately and validated through 5 replications.
It is possible to generate results within short period (few time units) through
simulation whereas it is tedious to obtain the results analytically.

6.7.1 Threshold Value Vs Average Components in the Queue

A comparative analysis between theoretical and simulated value is

performed in the Table 6.1. It presents the average number of components in
the queue for various threshold values of „a‟ and arrival rate λ. It can be
observed that the queue length (both theoretical and simulation) increases
when the threshold value increases. Similarly, the queue length increases
when the arrival rate increases. As seen in the Table 1, there is a good match
between the theoretical and simulation results. Then the variation is measured
by the relative error,

Theoretical value − Simulated value

Variation =
Theoretical value

The variation lies between 0.0011 - 0.1001 which is acceptable and

the results indicate that there is no significant variation in both the theoretical
and simulated values.

Figure 6.3 illustrates the effects of average components in the

queue for different values of threshold value. When the threshold value
increases then the average component in the queue (both theoretical and
 189

simulation) also increases. Also, it is clear from the Figure that there is a
slight variation between theoretical and simulation values which is negligible.

6.7.2 Arrival Rate and Service Rate vs Average Components in the

Queue

Table 6.2 shows the effects of average components in the queue for
various arrival rates (λ) and service rates (μ). As the arrival rate increases, the
average component in the queue (both theoretical and simulation) increases
for all service rates. Also it can be noted that the queue length decreases on
increasing the service rate.

Figure 6.4 depicts the effects of average component (obtained by

simulation) on increasing the arrival rate and service rate which is given in
Table 6.2. The average component shows the increasing trend on increasing
the arrival rate and decreasing trend on increasing the service rate.

6.7.3 Probability of Allowing Components During Service (α) Vs

Performance Metrics

Table 6.3 is constructed to show the effects of probability of

allowing ring cops/ components during service (α) on average number of
components in the queue, processing time and idle time. From the Table 6.3,
it can be seen that on increasing the probability of allowing components
during service (α), there is an increasing trend in the queue length. It can also
be observed that the processing time (idle time) of the server increases
(decreases) for increasing values of α.

Figure 6.5 illustrates the effects of probability of allowing

components during the busy period on the average components in the queue
 190

and processing time. When α value increases then the processing time as well
as average component increases.

6.7.4 Simulation Results of Different Objects /States

The different parameters like average number of components,

average staytime, idle time, processing time, inputs and outputs for different
objects/states at an arbitrary time are summarized in the Tables 6.4 (a), 6.4 (b)
and 6.5. These performance metrics are found for various arrival rates,
service rates and threshold values using simulation.

6.7.4.1 For various arrival and service rates

The parameters like average number of components, staytime, etc.,

are found for different arrival rates and service rates. From the Table 6.4 (a)
and 6.4 (b), the following observations are noticed:

o When the arrival rate (λ) of the components/ ring cops increases

 queue: the average number of components and average staytime

in the queue increases.

 exit1 and exit2: the average number of components in exit1 and

exit2 has no significant difference (slight variations due to
random number generation).

 server: the average number of components in the server has no

change, the average staytime in the server decreases slowly (no
significant difference) and idle time of the server decreases
whereas processing time increases.

o When the service rate (µ) of the server increases, the following
effects are seen:
 191

 queue: the average number of components and average staytime

in the queue decreases.

 exit1 and exit2: the average number of components in exit1

increases (slightly) whereas the average number of components in
exit2 has no significant difference.

 server: the average staytime in the server decreases and idle time
of the server increases whereas the processing time decreases.

Figure 6.6 is plotted for the processing time against arrival rate (λ)
for various values of service rate (μ). It depicts that the processing time
increases when the arrival rate increases whereas it decreases when the
service rate increases.

6.7.4.2 For various threshold value „a‟

The different parameters including the total number of inputs and

output of every object at an arbitrary time is noted in the Table 6.5. From the
Table 6.5, it is noted that when the threshold value increases, then

 source: the components from the source (which is used to

represent arrival of components) has no significant difference.

 queue: the average number of components and average staytime

in the queue increases.

 exit1 and exit2: the average number of components in exit1 and

exit2 has no significant difference.

 server: average staytime in the server increases slightly, the

average number of components and idle time of the server
increases whereas processing time decreases.
 192

Flexsim summary report

Time in minutes and components in units (1 unit=6 ring cops)

Table 6.1 Threshold value (a) Vs Average components in the queue

(components in units (1 unit=6 ring cops))
(For ξ=2.0, β=1.0, µ=2.0, α=0.8, b=10)

Arrival E(Q)/Average components in the queue

a
rate λ Theoretical Simulation Variation
2 1.884 1.886 .0011
3 2.268 2.118 .0661
4 2.695 2.481 .0794
5 3.165 2.927 .0752
λ=2.0

6 3.668 3.343 .0886

7 4.220 3.897 .0765
8 4.868 4.383 .0996
9 5.525 5.040 .0878
10 6.336 5.839 .0784
2 2.267 2.374 .0472
3 2.667 2.571 .0360
4 3.108 2.881 .0730
5 3.591 3.269 .0897
λ=2.5

6 4.112 3.710 .0978

7 4.667 4.299 .0789
8 5.294 4.772 .0986
9 5.994 5.411 .0973
10 6.829 6.185 .0943
2 2.621 2.850 .0874
3 3.080 3.051 .0094
4 3.534 3.379 .0439
5 4.027 3.671 .0884
λ=3.0

6 4.566 4.156 .0898

7 5.146 4.665 .0935
8 5.784 5.205 .1001
9 6.512 5.879 .0972
10 7.340 6.710 .0858

It can also be noted that the threshold value doesn‟t affect the total
number of inputs and outputs of the queue (only slight variations), server,
 193

exit1 and exit2. Figure 6.7 shows the effects of threshold value (a) of
components/ ring cops on processing and idle time of the server. The
processing time decreases and idle time increases when the threshold value
increases.

Table 6.2 Arrival rate (λ) and Service rate (µ) Vs Average components in
the queue
(For a=3, b=4, α=0.8, β=1.0, ξ=2.0)

Service rate Arrival rate λ Average components in the queue

Theoretical Simulation
1.0 1.812 1.711
1.5 2.414 2.464
µ=2.0

2.0 3.143 3.852

2.5 4.059 6.717
3.0 5.323 7.101
1.0 1.799 1.627
1.5 2.331 2.140
µ=2.5

2.0 2.936 2.903

2.5 3.618 4.052
3.0 4.435 6.280
1.0 1.789 1.582
1.5 2.298 2.001
µ=3.0

2.0 2.839 2.518

2.5 3.447 3.271
3.0 4.101 4.405
1.0 1.779 1.551
1.5 2.283 1.917
µ=3.5

2.0 2.805 2.363

2.5 3.371 3.010
3.0 3.951 3.761
 194

Table 6.3 Probability of allowing components during busy period (α) Vs

Performance metrics
(For a=3, b=4, ξ=2.0, β=1.0, µ=2.0, λ=0.5)
Average components in Processing Idle time
α
the queue time
0.2 1.055 5125.599 104374.90
0.4 1.087 9998.32 99052.25
0.6 1.150 14798.55 93876.85
0.8 1.248 19520.85 88859.70
1.0 1.345 23633.01 84521.56

Table 6.4 (a) Arrival rate and Service rate μ=2.5 Vs Performance metrics
(For a=3, b=4, α=0.8, β=1.0, ξ=2.0)

μ=2.5
Arrival Object
rate Average Average Idle Processing
components staytime time time
Source - - - -
Queue 1.627297 1.042316 - -
λ=1.0 Exit1 1.12877 - - -
Server 3.5 0.87523 61851.75 33977.82
Exit2 0.999973 - - -
Source - - - -
Queue 2.139571 1.186083 - -
λ=1.5 Exit1 1.121659 - - -
Server 3.5 0.861244 58025.928 40530.442
Exit2 0.999966 - - -
Source - - - -
Queue 2.903448 1.206744 - -
λ=2.0 Exit1 1.124346 - - -
Server 3.5 0.845487 47522.121 54418.152
Exit2 0.999985 - - -
 195

Table 6.4 (a) (Continued)

Source - - - -
Queue 4.051664 1.354333 - -
λ=2.5 Exit1 1.127515 - - -
Server 3.5 0.83213 34624.877 64149.256
Exit2 0.999985 - - -
Source - - - -
Queue 6.279874 1.749873 - -
λ=3.0 Exit1 1.123529 - - -
Server 3.5 0.822204 28055.588 90756.224
Exit2 0.999979 - - -
Source-Arrival of components; Exit1-Number of components exit from the system due
to limited number of admissions; Exit2-Number of components exit from the system
after service completion; Nil value in a row represents Not Applicable/Null
 196

Table 6.4 (b) Arrival rate and Service rate μ=3.5 Vs Performance metrics
(For a=3, b=4, α=0.8, β=1.0, ξ=2.0)
μ=3.5
Arrival
Object Average Average Idle Processing
rate
components staytime time time

Source - - - -
Queue 1.551313 0.685627 - -
λ=1.0 Exit1 1.135183 - - -
Server 3.5 0.64835 99382.554 17422.8549
Exit2 0.999934 - - -
Source - - - -
Queue 1.916659 0.812792 - -
λ=1.5 Exit1 1.126247 - - -
Server 3.5 0.639585 63673.868 26921.12624
Exit2 0.999977 - - -
Source - - - -
Queue 2.363245 0.908606 - -
λ=2.0 Exit1 1.126531 - - -
Server 3.5 0.628242 74127.818 46975.29605
Exit2 0.999982 - - -
Source - - - -
Queue 3.010476 0.977877 - -
λ=2.5 Exit1 1.123532 - - -
Server 3.5 0.619022 50727.045 46701.52979
Exit2 0.999987 - - -
Source - - - -
Queue 3.761477 1.018018 - -
λ=3.0 Exit1 1.1245 - - -
Server 3.5 0.609291 47142.111 61757.52634
Exit2 0.999981 - - -
Source-Arrival of components; Exit1-Number of components exit from the system due
to limited number of admissions; Exit2-Number of components exit from the system
after service completion; Nil value in a row represents Not Applicable/Null
 197

Table 6.5 Threshold value (a) Vs Performance metrics

(For ξ=2.0, β=1.0, µ=2.0, α=0.8, b=10, λ=2.0)

Total no. of Output

Total no. of Input

components
Components

components
Processing
Staytime
Average
Average
Object

Time
Time
Idle
a

2 Source - - - - - 309846
Queue 1.88601 0.774156 - - 247967 247967
Exit1 1.124699 - - - 61879 -
Server 6.00 1.047884 38500.62 61933.4 247967 247967
Exit2 0.999983 - - - 247967 -
3 Source - - - - - 310402
Queue 2.11808 0.879221 - - 248287 248283
Exit1 1.126021 - - - 62115 -
Server 6.50 1.059764 49455.45 50774.12 248283 248283
Exit2 0.99997 - - - 248283 -
4 Source - - - - - 310039
Queue 2.481296 1.026926 - - 248123 248123
Exit1 1.122683 - - - 61916 -
Server 7.00 1.067107 57350.9 42893.31 248123 248123
Exit2 0.999978 - - - 248123 -
5 Source - - - - - 311513
Queue 2.927117 1.191185 - - 249408 249408
Exit1 1.12276 - - - 62105 -
Server 7.50 1.079088 62719.14 37664.77 249408 249408
Exit2 0.999977 - - - 249408 -
6 Source - - - - - 308942
Queue 3.343402 1.392999 - - 247261 247261
Exit1 1.125223 - - - 61681 -
Server 8.00 1.084975 67437.39 33119.57 247261 247261
Exit2 0.999957 - - - 247261 -
 198

Table 6.5 (Continued)

Total no. of Output

Total no. of Input

components
Components

components
Processing
Staytime
Average

Average
Object

Time
Time
Idle
a

7 Source - - - - - 309549
Queue 3.897214 1.603049 - - 247521 247521
Exit1 1.123338 - - - 62028 -
Server 8.50 1.087381 70911.3 29814.65 247521 247520
Exit2 0.999959 - - - 247520 -
8 Source - - - - - 310287
Queue 4.383268 1.826062 - - 248166 248159
Exit1 1.124918 - - - 62121 -
Server 9.00 1.093522 73235.16 27619.71 248159 248159
Exit2 0.999957 - - - 248159 -
9 Source - - - - - 309208
Queue 5.040098 2.105447 - - 247186 247178
Exit1 1.124461 - - - 62022 -
Server 9.50 1.088454 75312.85 25666.88 247178 247178
Exit2 0.999959 - - - 247178 -
10 Source - - - - - 309852
Queue 5.839133 2.410098 - - 247694 247690
Exit1 1.12432 - - - 62158 -
Server 10.00 1.087254 76329.81 24715.91 247690 247690
Exit2 0.999959 - - - 247690 -
Source-Arrival of components; Exit1-Number of components exit from the system due
to limited number of admissions; Exit2-Number of components exit from the system
after service completion; Input-Total input to the object at an arbitrary time; Output-
Total output from the object at an arbitrary time; Nil value in a row represents Not
Applicable/No value
 199

Figure 6.3 Threshold value (a) Vs Average components in the queue

(For ξ=2.0, β=1.0, µ=2.0, α=0.8, b=10, λ=2.0)

8
µ=2.0
7
µ=2.5
6 µ=3.0
5 µ=3.5
Average components

1
1 1.5 2 2.5 3
Arrival rate (λ)

Figure 6.4 Arrival rate Vs Average components in queue

(For a=3, b=4, α=0.8, β=1.0, ξ=2.0)
 200

Figure 6.5 Probability of allowing components during busy period (α) Vs

Processing Time Vs Average components
(For a=3, b=4, ξ=2.0, β=1.0, µ=2.0, λ=0.5)

μ=3.5
100000
Processing Time (mins)

μ=2.5
80000

60000
40000
20000
0 μ=2.5
1 μ=3.5
1.5
2
2.5
Arrival rate (λ) 3

Figure 6.6 Arrival rate and Service rate Vs Processing Time

(For a=3, b=4, α=0.8, β=1.0, ξ=2.0)
 201

Figure 6.7 Threshold value (a) Vs Processing and Idle time

(For ξ=2.0, β=1.0, µ=2.0, α=0.8, b=10, λ=2.0)

6.8 VERIFICATION AND VALIDATION OF THE

SIMULATION MODEL

Verification is the process of evaluating whether the model

satisfies the conditions imposed at the start of that phase and works as
planned. Based on the results observed in the above tables, it is clear that the
model provides results as planned and it is verified that the process step is
similar between the simulated model and theoretical result.

Validation is the process of evaluating the model during or at the

end of the execution process to determine whether it satisfies the specified
expectations and requirements. In every simulation model, it is necessary to
validate the model by comparing the results with any known analytical
results. Compiler in the Flexsim program has been used to check that the
 202

model runs without exceptions/errors. It is clear that there are no flexscript

exceptions caught while executing the model. For validation purpose, the
model has been simulated 5 times (5 replications). In order to know the
validity of the model, it is compared with the theoretical results. The
variation between the analytical and simulated model lies between 0.0011 -
0.1001 which is acceptable, therefore, it can be concluded that the developed
model is valid.

Since analytical method is a tedious process to obtain various

performance metrics, simulation is performed.

6.9 CONCLUSION

In this chapter, simulation of M X / G(a, b) /1 queueing system with

limited number of admissions and multiple vacations has been discussed
through an industrial application. Numerical illustration is provided using
simulation.