You are on page 1of 18

Objective

Our objective in this course work is to simulate the arrival and service
pattern of customers at a branch of a post office. Here we got total three counter
services that will be used to serve the customer. The pattern of service will be FCFS
(First Come First Served) as queue discipline method. Our goal here is to do a
decision analysis whether to start using all three counters from the opening of the
branch or to use the selected counters in regards of the customer arrival pattern.
In our simulation procedure we will assume that, there will be only one
single queue no matter how many counter services is operating there in a particular
period of time (9am to 12 pm in our case study).

We have some data that are pre-determined which is shown below:

Table 1: Customer arrival rate

Time between Arrivals Probability


(min)
1 0.35
2 0.46
3 0.12
4 0.05
5 0.02

Table 2: Service time Duration

Duration of service Probability


time(min)
3 0.15
4 0.23
5 0.50
6 0.12

Table 3: Queue length

Number in Queue (n*) Probability


5≤n≤7 0.20
8≤n≤10 0.40
n>10 0.50
*n=number of customer in the queue

__________________________________________________________________
Page 1 of 18
Copyright © 2008 M R I Bhuiyan; London Metropolitan University
Introduction

1.1 Basic concept of queuing


The basic meaning of queue represents a number of customers waiting in line to
be served by the server. This is how all of us will generate the simple idea about a
queue. In practice you will get a handful of people who dose not queue up in their day
to day life. This queue may be passengers waiting for bus at the bus stop, customers
waiting to be served at any super store or bank or any food store etc.

1.2 How the queue is created


Consider yourself in a super market where you have finished your shopping and
ready to pay and go home. You head towards the server and found that all the servers
are busy at that time being with other customers. So definitely no server can serve you
until and unless any of the servers finished with their customers they are dealing with.
So you lined up to be served by forming a queue. In this case you may be the
originator of the queue or may in practice found yourself as part of the queue where
some other customers are waiting as well before you.

1.3 Elements of a queue


Let us consider the example that is been stated above (1.2). There are three basic
elements of a queue.
1) The customer
2) The server and
3) The queue

The customer (not mandatory to be a person) will continue to come over but the main
consideration here is how the customer is coming, the flow of the customer. Note that
in queuing system it’s not mandatory for a customer to be a person. Customer
behaviour like how the arrive is key factor in a queue system.

The server is what that provides service to the customer whether there is a queue or
not. The service mechanism of a server is our main consideration in a queuing system.
The efficiency of a server can help reduce the queue length in several ways.

A group of customers that have requested/waiting for service but haven’t yet received
any service forms a queue. Bear in mind queue discipline is a considerable factor, on
what basis these waiting customers will be served, will it be FCFS (First Come First
Served) or LCFS (Last Come First Served). Next considerable factor is whether there
is a single queue for the customers or more than one queue feeding the servers.

__________________________________________________________________
Page 2 of 18
Copyright © 2008 M R I Bhuiyan; London Metropolitan University
1.4 Types of queuing system
At any situation there can be three types of queuing system. But the element of
the queuing system is pre determined. The system efficiency depends on this elements
behaviour. A queuing system can be more effective and efficient if the elements of the
system are organized and it is known to everyone how the system will behave at a
particular time period of the day. Depending on the queue behaviour and service
pattern a system will work sound. There can be three types of queuing system;
i) single server single queue
ii) parallel server single queue
iii) parallel server more than one queue

In practice Single Server Single Queue system is not that much efficient and effective
for a busy and bigger organization .It can create too much balking of customers
depending on the arrival rate of the customer making the system much more cost
effective. Whereas in parallel server system more than one server works at a time
reducing the system cost and customer balking.

1.5 Some basic terms of queuing system:

Arrival time: the time that the customer arrives at the queue.

Departure time: the time that the customer gets out of the system after successfully
completing the service. Departure time can be measured in two way; departure time
from the queue and departure time from the system.

Waiting time: the time that the customer waits in the queue before getting served by
any server. It may also be called time in the queue.

Service time: the time it takes to serve a customer.

Time in the system: the total time that the customer has spent from joining the queue
and getting out of the system after getting required service.

Queue length: maximum number of customers that is in the queue waiting to be


served.

Balking: any customer who is not joining the queue upon arrival and leaving the
system without being server. The more the balking rate is the less the system
efficiency is.

__________________________________________________________________
Page 3 of 18
Copyright © 2008 M R I Bhuiyan; London Metropolitan University
Single Server Single Queue

2.1 Overview of a single server single queue system


A single server single queue system is one where there is only one server serving
in the whole system and only one customer queue feeding the server. In practice this
kind of queuing system is not that much effective to make a system working properly.
There is a fare chance that this kind of system proposal will create chaos in the system
and thus make the whole system a useless one. But however there is some certain
probability that this kind of system may be adopted by certain institutions where the
customer flow is not that much high and predetermined. We will see in our case study
how a single server system can be useless by making the total system less effective
and increase the waiting cost.

2.2 Worked Example of a single server single queue system


Based on our case study criteria the following table (created using Microsoft
Excel) shows the outcome of the system where there is only one server and one queue
feeding the system.
NB: the random number generation sheet is being provided in different page.

Post office system analysis using single server single queue

Number
time between Probability Cumulative in Probability Cumulative
arrivals(min) probability Queue(n) probability
1 0.35 0.35 5<=n<=7 0.2 0.2
2 0.46 0.81 8<=n<=10 0.4 0.6
3 0.12 0.93 n>10 0.5 1.1
4 0.05 0.98
5 0.02 1

Duration of Probability Cumulative


service
time(min) Probability
3 0.15 0.15
4 0.23 0.38
5 0.5 0.88
6 0.12 1

Discrete random number generation

Service No of
Inter arrival Arrival start waiting Service Completion Time in the customer

Customer time(min) time time Time(min) time(min) time System in the system
1 00:01 08:58 09:00 00:02 00:04 09:04 00:06 1

__________________________________________________________________
Page 4 of 18
Copyright © 2008 M R I Bhuiyan; London Metropolitan University
2 00:01 08:59 09:04 00:05 00:04 09:08 00:09 2
3 00:03 09:02 09:08 00:06 00:05 09:13 00:11 3
4 00:01 09:03 09:13 00:10 00:03 09:16 00:13 4
5 00:03 09:06 09:16 00:10 00:05 09:21 00:15 4
6 00:05 09:11 09:21 00:10 00:05 09:26 00:15 5
7 00:03 09:14 09:26 00:12 00:05 09:31 00:17 5
8 00:03 09:17 09:31 00:14 00:05 09:36 00:19 5
9 00:01 09:18 09:36 00:18 00:05 09:41 00:23 6
10 00:02 09:20 09:41 00:21 00:05 09:46 00:26 7
11 00:03 09:23 09:46 00:23 00:05 09:51 00:28 7
12 00:01 09:24 09:51 00:27 00:04 09:55 00:31 8
13 00:03 09:27 09:55 00:28 00:05 10:00 00:33 8
14 00:02 09:29 10:00 00:31 00:05 10:05 00:36 9
15 00:03 09:32 10:05 00:33 00:04 10:09 00:37 9
16 00:01 09:33 10:09 00:36 00:03 10:12 00:39 10
17 00:02 09:35 10:12 00:37 00:05 10:17 00:42 11
18 00:02 09:37 10:17 00:40 00:05 10:22 00:45 11
19 00:04 09:41 10:22 00:41 00:03 10:25 00:44 11
20 00:03 09:44 10:25 00:41 00:05 10:30 00:46 12
21 00:02 09:46 10:30 00:44 00:05 10:35 00:49 12
22 00:01 09:47 10:35 00:48 00:05 10:40 00:53 13
23 00:02 09:49 10:40 00:51 00:06 10:46 00:57 14
24 00:01 09:50 10:46 00:56 00:05 10:51 01:01 15
25 00:02 09:52 10:51 00:59 00:04 10:55 01:03 15
26 00:01 09:53 10:55 01:02 00:05 11:00 01:07 16
27 00:01 09:54 11:00 01:06 00:05 11:05 01:11 17
28 00:02 09:56 11:05 01:09 00:05 11:10 01:14 17
29 00:02 09:58 11:10 01:12 00:05 11:15 01:17 18
30 00:02 10:00 11:15 01:15 00:04 11:19 01:19 18
31 00:01 10:01 11:19 01:18 00:05 11:24 01:23 19
32 00:03 10:04 11:24 01:20 00:04 11:28 01:24 20
33 00:02 10:06 11:28 01:22 00:04 11:32 01:26 20
34 00:01 10:07 11:32 01:25 00:05 11:37 01:30 21
35 00:02 10:09 11:37 01:28 00:05 11:42 01:33 21
36 00:03 10:12 11:42 01:30 00:05 11:47 01:35 21
37 00:02 10:14 11:47 01:33 00:04 11:51 01:37 22
38 00:01 10:15 11:51 01:36 00:05 11:56 01:41 23
39 00:01 10:16 11:56 01:40 00:04 12:00 01:44 24
40 00:04 10:20 12:00 01:40 00:06 12:06 01:46 24
41 00:03 10:23 12:06 01:43 00:05 12:11 01:48 24
42 00:01 10:24 12:11 01:47 00:05 12:16 01:52 25
43 00:04 10:28 12:16 01:48 00:04 12:20 01:52 25
44 00:01 10:29 12:20 01:51 00:04 12:24 01:55 26
45 00:02 10:31 12:24 01:53 00:06 12:30 01:59 26
46 00:02 10:33 12:30 01:57 00:06 12:36 02:03 27
47 00:03 10:36 12:36 02:00 00:04 12:40 02:04 27
48 00:01 10:37 12:40 02:03 00:05 12:45 02:08 28
49 00:02 10:39 12:45 02:06 00:05 12:50 02:11 29
50 00:03 10:42 12:50 02:08 00:05 12:55 02:13 29
51 00:02 10:44 12:55 02:11 00:06 13:01 02:17 30
52 00:03 10:47 13:01 02:14 00:05 13:06 02:19 30
53 00:02 10:49 13:06 02:17 00:05 13:11 02:22 31

__________________________________________________________________
Page 5 of 18
Copyright © 2008 M R I Bhuiyan; London Metropolitan University
54 00:01 10:50 13:11 02:21 00:03 13:14 02:24 32
55 00:01 10:51 13:14 02:23 00:03 13:17 02:26 32
56 00:01 10:52 13:17 02:25 00:04 13:21 02:29 33
57 00:02 10:54 13:21 02:27 00:04 13:25 02:31 34
58 00:01 10:55 13:25 02:30 00:04 13:29 02:34 34
59 00:01 10:56 13:29 02:33 00:06 13:35 02:39 35
60 00:02 10:58 13:35 02:37 00:03 13:38 02:40 36
61 00:01 10:59 13:38 02:39 00:04 13:42 02:43 37
62 00:02 11:01 13:42 02:41 00:04 13:46 02:45 37
63 00:03 11:04 13:46 02:42 00:04 13:50 02:46 38
64 00:01 11:05 13:50 02:45 00:05 13:55 02:50 38
65 00:02 11:07 13:55 02:48 00:05 14:00 02:53 39
66 00:01 11:08 14:00 02:52 00:03 14:03 02:55 40
67 00:01 11:09 14:03 02:54 00:04 14:07 02:58 41
68 00:01 11:10 14:07 02:57 00:05 14:12 03:02 41
69 00:02 11:12 14:12 03:00 00:05 14:17 03:05 42
70 00:02 11:14 14:17 03:03 00:05 14:22 03:08 43
71 00:02 11:16 14:22 03:06 00:05 14:27 03:11 43
72 00:01 11:17 14:27 03:10 00:06 14:33 03:16 44
73 00:01 11:18 14:33 03:15 00:04 14:37 03:19 45
74 00:01 11:19 14:37 03:18 00:04 14:41 03:22 45
75 00:01 11:20 14:41 03:21 00:03 14:44 03:24 46
76 00:02 11:22 14:44 03:22 00:05 14:49 03:27 47
77 00:01 11:23 14:49 03:26 00:06 14:55 03:32 48
78 00:02 11:25 14:55 03:30 00:05 15:00 03:35 48
79 00:02 11:27 15:00 03:33 00:03 15:03 03:36 49
80 00:01 11:28 15:03 03:35 00:05 15:08 03:40 49
81 00:02 11:30 15:08 03:38 00:05 15:13 03:43 50
82 00:01 11:31 15:13 03:42 00:05 15:18 03:47 51
83 00:01 11:32 15:18 03:46 00:06 15:24 03:52 51
84 00:01 11:33 15:24 03:51 00:05 15:29 03:56 52
85 00:03 11:36 15:29 03:53 00:06 15:35 03:59 53
86 00:01 11:37 15:35 03:58 00:05 15:40 04:03 53
87 00:01 11:38 15:40 04:02 00:05 15:45 04:07 54
88 00:04 11:42 15:45 04:03 00:05 15:50 04:08 54
89 00:02 11:44 15:50 04:06 00:05 15:55 04:11 55
90 00:02 11:46 15:55 04:09 00:05 16:00 04:14 56
91 00:01 11:47 16:00 04:13 00:05 16:05 04:18 56
92 00:02 11:49 16:05 04:16 00:04 16:09 04:20 57
93 00:01 11:50 16:09 04:19 00:05 16:14 04:24 58
94 00:02 11:52 16:14 04:22 00:05 16:19 04:27 58
95 00:02 11:54 16:19 04:25 00:03 16:22 04:28 59
96 00:01 11:55 16:22 04:27 00:06 16:28 04:33 60
97 00:02 11:57 16:28 04:31 00:06 16:34 04:37 60
98 00:05 12:02 16:34 04:32 00:04 16:38 04:36 60
99 00:01 12:03 16:38 04:35 00:05 16:43 04:40 61
100 00:02 12:05 16:43 04:38 00:06 16:49 04:44 62

__________________________________________________________________
Page 6 of 18
Copyright © 2008 M R I Bhuiyan; London Metropolitan University
The arrival time of first customer is assumed at 8:58 am. The post office opens at
9:00 am. Assuming that the server will start service as soon as the post office opens
and the customers are served in FCFS method.

So waiting time for the first customer = service start time – arrival time
=9:00-8:58
=2 min(s)

Service time for the first customer = 4 minutes

Therefore the first customer is in the system till 9:04 am. In the meantime if there is
any arrival of other customer they will have to queue up for getting service until the
server finished with the customer getting service [as shown in figure below].

Service Time in No of
Inter arrival Arrival start waiting Service Completion the customer

Customer time(min) time time Time(min) time(min) time System in the system
1 00:01 08:58 09:00 00:02 00:04 09:04 00:06 1
2 00:01 08:59 09:04 00:05 00:04 09:08 00:09 2

Fig 1: customer in the queue

Total time in the system for each customer = waiting time + service time

For the first customer total time in the system = 2 min(s) + 4 min(s)
= 6 min(s)
For the second customer total time in the system = 5 min(s) + 4 min (s)
= 9 min(s)
For nth customer total time in system = nth waiting time + nth service time

Here inter arrival time and service time was generated by random number generation
process using Microsoft Excel.

2.3 Outcomes discussion on system performance

From the above experiment we can measure some possible outcome of the system.
How the system will work, how effective its cost will be, how this type of system will
effect the customer creating more balking more waiting cost and at the end totally
useless.

I have counted 100 customer data arriving within the time period 09:00 am to 12:00
pm (in practice12:05 pm). When the customer arrived? when the customer started
getting service? How long the customer had to wait for getting service?
In this section I will try to figure out the system efficiency by generating step by step
calculation.
Cell A represents the customer number
Cell B represents inter arrival time [generated by random number generation]
Cell C represents Customer arrival time in hh:mm format
Cell D represents service start time for each customer
Cell E represents waiting time for each customer
__________________________________________________________________
Page 7 of 18
Copyright © 2008 M R I Bhuiyan; London Metropolitan University
Cell F represents how long the customer is being served [generated by random
number generation]
Cell G represents service completion time
Cell H represents total time in the system for each customer in the system
Cell I represents no of customer in the system for a time period

As we can see from the example above the first customer who arrived at 8:58 am is
being served at 9:00 am, the customer had to wait 2 minutes to be served. The
customer had got 4 minutes of service hence getting out of the system at 9:04 am. The
second customer arrived at 8:59 am and was waiting to be served. As soon as the first
customer was served the second customer started to be getting served. Hence the
waiting time for the second customer becomes 5 minutes. So by 9:04 there are already
3 customers in the system as one more customer has arrived at 9:03.The same
procedure occurred for all customers.

Here the service time is much greater than that of the customer arrival rate. This is a
reason that is making the customer to wait more time in the system and the potential
risk of customer bulking is increasing. But we will assume that all the customer who
arrives between 9:00 am to 12:00 pm will get service.

Hence the customer bulking rate for this period of time is 0 %.

The average Interarrival time (1/λ) = 1


So, the average arrival rate of customers λ =1

Interarrival time(min)

Interarrival time(min)

00:05
00:05
00:04
00:03
00:02
00:02
00:01
00:00
00:00
1
5
9
13
17
21
25
29
33
37
41
45
49
53
57
61
65
69
73
77
81
85
89
93
97

Fig 2: Customer Interarrival time

The average service time (1/μ) = 4 min(s)


So the average rate at which the server serves the customer μ =1/4 = 0.25

Hence the traffic intensity ρ = λ/μ


= 1/0.25
=4
The value of ρ > 1 means the system can not reach a steady state. The value of ρ here
is 4 so at least 4 servers will be needed to make the system steady.
__________________________________________________________________
Page 8 of 18
Copyright © 2008 M R I Bhuiyan; London Metropolitan University
A possible outcome of this system with 4 servers using WinQsb has been shown
in Appendix.

As the value of ρ > 1, it is usual to assume that the possible idle time of each cashier
in this system is 0.

The average number of customer in the system is 31.


The average time a customer spent in the system 2 hours 17 minutes

The graphical representation for first 30 customer’s arrival and how many
customers are there already in the system given below shows how the number of
customer in the system increasing as the time increasing.

Arrival No of custmer

20
18
customer in the system

16
14
12
10
8
6
4
2
0
9

0
e

:5

:0

:1

:1

:2

:2

:2

:3

:3

:4

:4

:5

:5

:5

:0
t im

08

09

09

09

09

09

09

09

09

09

09

09

09

09

10
Fig 3: number of customer in the system

From the above graph it is clearly visible that as the time is going on the number
of customer in the system is increasing because the arrival rate is much higher than
that of the server’s service rate.

In figure 2 and 3 we can see that none of the graph is showing any kind of normal
distribution properties.

2.4 Opportunities for this system

In this system we are using only one server that is the main reason why the system
is not steady and no idle time for cashier. Though we have omitted customer balking
rate but in practice this system not useable. If we install more servers in this system
the system will behave normally making it a good one and customer friendly. A
possible result is shown through WinQsb for 4 servers.

__________________________________________________________________
Page 9 of 18
Copyright © 2008 M R I Bhuiyan; London Metropolitan University
Parallel Server Single Queue
3.1 Overview of parallel server system
The main concept of parallel server system is same as the single server system.
Waiting time, arrival time, queue length number of customer in the system everything
is calculated as the same manner as in single server system but the difference is in the
number of server working.
In this part I will evaluate the performance measurement of the post office system by
using 2 servers and 3 servers respectively. But instead of using Microsoft Excel I will
use WinQsb. The basic notation for this kind of queuing technique will be

A/B/n/N
This kind of notation is called Kendall’s notation for queues
Where
A is the distribution of Interarrival times.
B is the distribution of service times.
n is the number of servers.
N is the maximum number of customer allowed in the system at any one time.
For example if N=15 it means any customer arriving when there is 15 customer
already in the system may have to balk without receiving any service.

3.2 Performance measurement of 2 server single queue

Given data:

Customer arrival rate

Time between Arrivals Probability


(min)
1 0.35
2 0.46
3 0.12
4 0.05
5 0.02

Mean = (1×0.35) + (2×0.46) + (3×0.12) + (4×0.05) + (5×0.02)


=0.35+0.92+0.36+0.20+0.10
=1.93

Service time Duration

Duration of service Probability


time(min)
3 0.15
4 0.23
5 0.50
6 0.12

__________________________________________________________________
Page 10 of 18
Copyright © 2008 M R I Bhuiyan; London Metropolitan University
Mean = (3×0.15) + (4×0.23) + (5×0.50) + (6×0.12)
=0.45+0.92+2.50+0.72
=4.59

The following values will be used in WinQsb to find out system performance
measurement

The average Interarrival time 1/λ= 1.93


So, λ = (1/1.93) = 0.5181

The average service time 1/μ =4.59


So, μ= (1/4.59) = 0.2178

We have 2 servers working so number of servers n=2

By putting these values of n, λ and μ into WinQsb simulation software the following
results was found.

Table 4: Performance Measure Result for 2 server system

1 .System: M/M/2/15 From Formula


2. Customer arrival rate (lambda) per min = 0.5181
3. Service rate per server (mu) per min = 0.2178
4. Overall system effective arrival rate per min = 0.4314
5. Overall system effective service rate per min = 0.4314
6. Overall system utilization = 99.0400 %
7. Average number of customers in the system (L) = 12.6048
8. Average number of customers in the queue (Lq) = 10.6240
9. Average number of customers in the queue for a busy system (Lb) = 10.7838
10. Average time customer spends in the system (W) = 29.2172 mins
11. Average time customer spends in the queue (Wq) = 24.6259 mins
12. Average time customer spends in the queue for a busy system (Wb) =24.9962
mins
13. The probability that all servers are idle (Po) = 0.4385 %
14. The probability an arriving customer waits (Pw or Pb) = 98.5185 %
15. Average number of customers being balked per min = 0.0867

__________________________________________________________________
Page 11 of 18
Copyright © 2008 M R I Bhuiyan; London Metropolitan University
Fig 4: Simulated outcomes for 2 servers

Fig 5: Probability distribution

Fig 6: graphical representation of the system performance

__________________________________________________________________
Page 12 of 18
Copyright © 2008 M R I Bhuiyan; London Metropolitan University
3.2.1 System analysis

The value of λ = 0.5181


The value of μ =0.2178
So ρ = λ/ μ
= (0.5181/0.2178)
= 2.37
For a system to be in steady state the arrival rate of customers must be less than the
combined service rate of the n servers , means λ<n μ or (ρ/n)<1

Here in this system, the system cannot reach the steady state because the value of
λ=0.5181 is greater than (nμ) = (2×0.2178) = 0.4356. Hence the system will be in
transient state.

The typical graph for a system to be in steady state is shown below.* The figure
below is taken from our tutors lecture (week 4-5)*

1
qt
0.8
probability

0.6

0.4 s te ady s tate

0.2

0
0 0.5 1 1.5 2
time t

Fig 7: A system in steady state

The average time a customer spends in the system = 29.2172 mins = 29 mins
(approx)

The average queue length = 10.6240 = 11 (approx)


The percentage of customers who balk = 0.0867 %
The percentage of idle time for all cashier = 0.4385 %

3.3 Performance measurement of 3 server single queue

Value of average arrival rate of customers λ and the average rate at which the server
serves customers μ will be same as it was in section 3.2, but now the number of
servers will be 3. the following table shows the outcome of the system using three
servers.

__________________________________________________________________
Page 13 of 18
Copyright © 2008 M R I Bhuiyan; London Metropolitan University
Table 5: Performance Measure Result for 3 server system

1. System: M/M/3/18 From Formula


2. Customer arrival rate (lambda) per min = 0.5181
3. Service rate per server (mu) per min = 0.2178
4. Overall system effective arrival rate per min = 0.5160
5. Overall system effective service rate per min = 0.5160
6. Overall system utilization = 78.9664 %
7. Average number of customers in the system (L) = 4.5290
8. Average number of customers in the queue (Lq) = 2.1601
9. Average number of customers in the queue for a busy system (Lb) = 3.4288
10. Average time customer spends in the system (W) = 8.7778 mins
11. Average time customer spends in the queue (Wq) = 4.1864 mins
12. Average time customer spends in the queue for a busy system (Wb) = 6.6453
mins
13. The probability that all servers are idle (Po) = 5.9603 %
14. The probability an arriving customer waits (Pw or Pb) = 62.9980 %
15. Average number of customers being balked per min = 0.0021

Fig 8: Simulated outcomes for 3 servers

__________________________________________________________________
Page 14 of 18
Copyright © 2008 M R I Bhuiyan; London Metropolitan University
Fig 9: Probability distribution

Fig 10: graphical representation of the system performance

__________________________________________________________________
Page 15 of 18
Copyright © 2008 M R I Bhuiyan; London Metropolitan University
3.3.2 System analysis

The value of λ = 0.5181


The value of μ =0.2178
So ρ = λ/ μ
= (0.5181/0.2178)
= 2.37
For a system to be in steady state the arrival rate of customers must be less than the
combined service rate of the n servers , means (λ<n μ) or (ρ/n)<1

Here in this system, the system can reach the steady state because the value of
λ=0.5181 is less than (nμ) = (3×0.2178) = 0.6534.

The average time a customer spends in the system = 8.7778 mins = 9 mins (approx)

The average queue length = 2.1601= 2 (approx)


The percentage of customers who balk = 0.0021%
The percentage of idle time for all cashier = 5.9603 %

__________________________________________________________________
Page 16 of 18
Copyright © 2008 M R I Bhuiyan; London Metropolitan University
Summery Report
4.1 Worked carried out
My objective was to find out a solution for this Post Office how many service
positions should be kept out for a smooth service. To find out a proper solution I have
worked on three different aspects. I was been provided by the probability data of
customer arrival rate and servers service rate. On the basis of this data I carried on my
work. In my work you will see I worked on using one server then two server and
finally three servers. Each time the outcomes were different showing how the system
will work.
By assuming that there is only one single queue will feed the servers the results were
generated. It is obvious that if I would have used more than one single queue the
outcomes may be different. But it is recommended that one single queue is better than
two or three queue to deal with customer queries and transactions.

For single server system the outcomes was generated by using Microsoft Excel,
where the provided data was used and possible outcomes was found by using
RAND() function and the overall system performs was evaluated after some
mathematical functions. In this section I assumed the first customer arrived at 8:58
am. I took simulated result for 100 customers.

I used the WinQsb simulation software for two server and three server system.

My findings will be discussed in farther section 4.2.

4.2 Results
If you use only one server in the system you are going to invite chaos for you. If
you have a glance on my carried out work you can see how the waiting time for each
customer is increasing and how the number of customers in the system is increasing
rapidly as the time goes by.
For example at 11:00 am there are already 37 customers in the system and you
can easily guess how the number is going to be at 12:00pm. Have a look it is already
62 at 12:00 pm.
In my experiment I omitted possible balking otherwise there would have been a
lot of customers who would have balked without receiving any service effecting the
post office reputation. So my recommendation for you is to omit the idea for only one
server.
The second experiment was using two servers. The results that I got were fine but
still not good enough. Overall system utilization becomes high average queue length
is approximately 11 and probability of an arriving customer to wait is 98%.
The third experiment was using three servers. Here the outcomes were much
more satisfactory than that of the previous two experiments. Customer queue length is
much less average time in the system for each customer is competitive for any kind of
system.

__________________________________________________________________
Page 17 of 18
Copyright © 2008 M R I Bhuiyan; London Metropolitan University
__________________________________________________________________
Page 18 of 18
Copyright © 2008 M R I Bhuiyan; London Metropolitan University

You might also like