COMSATS

UNIVERSITY ISLAMABAD

EEE-354: Telecommunication
Systems Engineering

Queueing Systems Analysis

 Utilization for M/M/1 Infinite Queue
Example 1:
Consider I/O system with one disk and one controller. If average
time required to service each request is 6msec, what is the
maximum request rate it can tolerate?
Solution:
Maximum will occur when 100% utilized. So
𝝀
U = 𝜆ℎ 𝝁
is often called

1 = 𝜆ℎ traffic Intensity

$
𝜆!"# =
h
$
= =167 requests per second.
6∗0.001
 Utilization for M/M/1
If 𝝆 > 𝟏, 𝒕𝒉𝒆𝒏 𝝀 > 𝝁(𝒂𝒓𝒓𝒊𝒗𝒂𝒍 𝒓𝒂𝒕𝒆 𝒊𝒔 𝒈𝒓𝒆𝒂𝒕𝒆𝒓 𝒕𝒉𝒂𝒏 𝒔𝒆𝒓𝒗𝒊𝒄𝒆 𝒓𝒂𝒕𝒆)
• Jobs arrive faster than it can be processed.
• Queue grow Infinity
• Unstable
• Must have 𝜌 < 1 for stability so U is never greater than 100.

• As utilization increases Beyond 85%,

queue rises sharply
• Corresponding sharp rise in response
time must be under 100% but often
lower
• Ex: OS CPU Scheduler often
has 60-80%Heuristic.

I didn’t mention that in the video but the average waiting time in the system(average response time) also rises
sharply beyond 85% system utilization.
 M/M/1 with Infinite Queue Analysis
Example 2:
Customers arrive at rate 4 per hour, mean service time is 6 minutes.
Step1: Calculate 𝜆 and 𝜇

𝜆=4 per hour , 𝜇=10 per hour , Interarrival times: 15mins

System utilization :
!
=0.4 =40% of the time the server is being used.
"

Probability of 0 customer in the system

𝑃! (System is empty)= 1- r =0.6 (60% of the time server is free)

Probability of 1 customer in the system

𝑃" (one customer )= r1(1- r) = (2/5)(3/5)=6/24=0.24 =24%

Probability of 2 customers in the queue

𝑃# (three customers )= r3(1- r) = (3/5)(2/5)3
=3.8%
 M/M/1 with Infinite Queue Analysis
Probability of 10 customers in the system
𝑃"! (ten customer ) = r10(1- r) =( 3/5)(2/5)10
=0.0062 %

Probability of 0 customer in the queue

P(Queue is empty) = 𝑃# (Queue is empty and one at service)+ 𝑃! (Queue is empty and
no customer in the system)
= 0.6+0.24
=0.84=> 84% of the time
Probability of two or more than two customers in the system
P(there are more than two people in the system )=1-P(there are less than two
people in the system )
=1-[𝑃! + 𝑃" ]

=0.16=>16% of the time.

 M/M/1 with Infinite Queue Analysis
Expected number of people in the system
Average number of people in system = r/(1- r) =0.6 customer

Expected number of people in the Queue

Average Queue Length = r2/(1- r) = 0.26 customer

Expected waiting time in the System

Average waiting time in System= 0.16 hour

Expected waiting time in the Queue

Average Time in queue = Average time in system – average service time .
= 0.16 – 0.1 = 0.06 hour
 M/M/1 with Infinite Queue Analysis
Arrival rate =4 /hour Arrival rate = 8/hour Arrival rate = 9.9/hour

Probability of 0 60% 20% 1%

customer in the system

Probability of 0 84% 36% 1.9%

customer in the queue

Expected number of 0.6 customer 4 customers 99 customers

people in the system

Expected number of 0.26 customer 3.2 customers 98 customers

people in the Queue

Average Waiting time in 0.16 hour 0.5 hour 10 hours

system

Average Waiting time in 0.06 hour 0.4 hour 9.9 hours

queue
 M/M/1 with Infinite Queue Analysis
Probability that the waiting time of a customer in the
system/queue exceeds t.

For System :
𝑃 𝑇 ≥ 𝑡 = 𝑒K LKM N

For Queue :
𝜆 K(LKM)N
𝑃 𝑇≥𝑡 = 𝑒
𝜇
 M/M/1 with Infinite Queue Analysis
Example 3:
Telephone calls arrive at a telephone booth according to a Poisson process with a
mean interarrival time of 9 minutes. The length of telephone call is assumed to be
exponentially distributed with mean 3 minutes. Find the Probability that the waiting
time in the system and queue is greater than 10 minutes

Solution:
Step 1: Model Identification
Since there is only one telephone booth, the number of service channels is one. Also,
since any number of customers can enter the booth, the capacity of the system is
infinity. Hence this problem comes under the model M/M/1/∞/∞/FCFS
Step 2:
𝝀=1/9 min
𝝁=1/3 min
Probability that the waiting time in the system is greater than 10 minutes
$ $
# %#& ∗$(
𝑃 𝑇 ≥ 10 = 𝑒
=0.10
 M/M/1 with Infinite Queue Analysis
Probability that the waiting time in the queue is greater than 10
minutes
1 ( $($ ∗$,
𝑃 𝑇 ≥ 10 = 𝑒 ) *
3
=0.03
 M/M/1 with Infinite Queue
Example 4: Time between requests is exponentially distributed with mean time
between requests of 8ms . Time to process is also distributed exponentially with
average service times of 5ms.
a. What is the average response time?
b. How much faster must the server be to halve this average response ?
Solution :
𝜆=0.125 per ms
𝜇=0.200 per ms

Solution a :
Average response time
! ! !
Average waiting time in system="#! = $.&$#$.!&' = $.(' =13.33 ms
Solution b :
13.33 ms/2 =6.665; Assume 𝜆 fixed change 𝜇
!
6.665= ; 𝜇=0.275. 37.5% Faster
" #!
 M/M/1/ K/∞
• Probability of having 0 customers in system
1- r
p𝑃,0 =
1 - r K +1
• Probability of having j customers in system

𝑃p- = (1 - r ) r j
j K +1
, j = 1, 2,...K
1- r

j
ælö
p j = ç ÷ 𝑃p,0 , j = 0,1, 2,...K
èµø
 M/M/1 with Finite Queue
• Server Utilization
(1 - r ) r (1 - r K )
E [U ] = 1 - p𝑃0, = 1 - =
1 - r K +1 1 - r K +1

• Blocking Probability

(1 - r ) r K
PB = p𝑃.K =
1 - r K +1
Probability that an arriving customer
finds the queue full (at state K)
 M/M/1 with Finite Queue
Expected system Length
K
(1 - r ) K
(1 - r ) r K d {r j }
E [ X ] = å jp𝑃-j = å jr j
= å =
j =0 1- r K +1
j =0 1- r K +1
j =0 dr
(1 - r ) r d ì K ü (1 - r ) r d ì
ï1 - r K +1
üï
= íå r ý = ý=
j
í
1- r K +1
d r î j =0 þ 1 - r K +1
d r ïî (1 - r ) ïþ
(1 - r ) r æ (1 - r K +1 ) - (1 - r ) ( K + 1) r K ö
= K +1 ç ÷=
1- r ç ( 1 - r ) 2 ÷
è ø
r æ1- r K K ö
= K +1 ç
- Kr ÷
1- r è 1- r ø
 M/M/1/k/∞
Expected waiting time in system
• From the formula for L, we can derive the formula for W, the
average delay of a customer by Little’s Formula W = L/l. Note that the
arrival rate l includes blocked customers.
• The fraction of arrivals that actually enter the system is 1-Pk.
Hence the actual arrival rate is leff = l(1-Pk) and hence the average
delay is

W = L/leff = L/[l(1-Pk)].

leff = lcarried ; l= loffered

 M/M/1/k/∞
• Expected queue length
𝑃!$
!

• Expected waiting time in queue

"
Q
!
𝑊!P =W-
#
L
 M/M/1/k/∞
Example 5:
Customers arrive at rate 8 per hour, mean service time is 6 minutes.
System capacity K=12.
Step1: Calculate 𝜆 and 𝜇
Probability of 0 customer in the system
($#*)
P(System is empty) =𝑃( = =0.21 (21% of the time server is free)
$#*#$%#

System utilization :
1- 𝑃( of the time the server is being used.

Probability of 1 customer in the system

(!#*)*
P(one customer ) 𝑃! = !#* !"
=0.16
 M/M/1/k/∞
Probability of 12 customers in the system
("&')',-
P(blocking) 𝑃"$ = =1.4%
"&',.

Probability of 0 customer in the queue

P(Queue is empty) = P(Queue is empty and no one at service) 𝑃# +P(Queue is empty
and one customer being served) 𝑃%
= 𝑃" + 𝑃! =0.37
 M/M/1/k/∞
Expected number of people in the system
' "&',-
Average number of people in system L = ( − K𝜌"$
"&',-/, "&'

Expected number of people in the Queue

Average Queue Length = 𝐿) =L-(l(1-P12)/𝜇)

Expected waiting time in the System

Average waiting time in system W=L/[l(1-P12)].

Expected waiting time in the Queue

"
Average Waiting time in queue 𝑊/ =W-
*
 Comparing M/M/1 with Infinite and Finite
capacity
Arrival rate = 8/hour Arrival rate = 8/hour
Infinite Queue Model Finite Queue Model k=12

Probability of 0 customer in the system 20% 21%

Probability of 0 customer in the queue 32% 38%

Expected number of people in the system 4 customers 3.2 customers

Expected number of people in the Queue 3.2 customers 2.4 customers

Average Waiting time in system 0.5 hour 0.42 hours

Average Waiting time in queue 0.4 hour 0.32 hour

 M/M/1 with Finite Queue Analysis
Arrival rate = 8/hour Arrival rate = 8/hour
Finite Queue Model k=6 Finite Queue Model k=12
Probability of 0 customer in the system 25% 21%

Probability of 0 customer in the queue 45% 38%

Expected number of people in the system 2.2 customers 3.2 customers

Expected number of people in the Queue 1.2 customers 2.4 customers

Average Waiting time in system 0.28 hour 0.42 hours

Average Waiting time in queue 0.18 hour 0.33 hour

Blocking Probability 6.6% 1.4%

 M/M/c
𝟏 𝝀
𝑃, = ∑𝒄 (45)4 (45)4
. 𝝆 = (utilization)
𝒏1𝟏 𝒄𝝁
𝟏. 𝒏!
.𝒄!(𝟏85)

("#)0
(P>c)𝜅 = 𝑃( => Erlang C
"!(&'#)
#)
L= + cρ
&'#
#)
𝐿* =
&'#
) &
W= +
"+(&'#) +
)
𝑊* =
"+(&'#)
 M/M/c
Example 6
How does the response time in Example 4 changes if we increase the number of
servers to 4.
Solution So the average
𝜆=0.125 per ms response time
𝜇=0.200 per ms decreases by
approximately
(.$-. 62.5% when we
𝜌= =0.156
(.-(∗/ increase the
𝟏 umber of server
𝑃( = ∑𝒏(𝟏(+∗-.#/0)2 (+∗(-.#/0))+ = 𝟎. 𝟓𝟑𝟑
𝟏1 𝒏!
1 𝟒!(𝟏5𝟎.𝟏𝟓𝟑) to 4
(4 ∗ 0.156)/
𝜅= ∗ 0.533 = 0.004
4! (1 − 0.156)
(.((/ $
W= + =5.0ms
/∗(.-(($#(.$.2) (.-(