Sei sulla pagina 1di 4

EE132B-HW Set #9

UCLA 2014 Fall

Prof. Izhak Rubin

Problem 1
Consider a communication system which queues arriving messages and then transmits them on a First-Come First-Served (FCFS) basis across a single outgoing link.
Messages arrive in accordance with a Poisson point process, and the message service
time is exponentially distributed.
. Calculate the mean queue size and
(a) Let the service rate be equal to 0.85 messages
sec
the mean message waiting time for three cases:
messages
.
sec
0.8 messages
.
sec
.
0.84 messages
sec

(i) Message arrival rate = 0.45


(ii) Message arrival rate =
(iii) Message arrival rate =

(b) For each of the three arrival rate cases considered in part (a) above, compute
the probability that a message will have to wait for the start of its transmission
for longer than 2.5 sec.
Ans:
(a)
Given message arrival rate and service rate , the mean queueing size (i.e.,
2
E[Q]) is equal to 1
, where = .

E[Q]

(i) = 0.45
0.5956

(ii) = 0.8
15.0588

(iii) = 0.84
83.0118

Table 1: Mean queue size


(b)
The probability that a message will have to wait for the start of its transmission
for longer than t sec is equal to P (W > t) = 1 W (t) = e(1)t , for t 0.

P (W > t)

(i) = 0.45
0.1948

(ii) = 0.8
0.8306

(iii) = 0.84
0.9638

Table 2: The probability of a message waiting for longer than 2.5 sec to start its
transmission.

Problem 2
Consider the following communications system. Messages arrive in accordance
with a Poisson point process at a transmission node that is served by a single transmission channel. Messages are queued in the nodal buffer until they reach the head
1

EE132B-HW Set #9

UCLA 2014 Fall

Prof. Izhak Rubin

of the queue, at which time the message is transmitted across the single outgoing
channel at a transmission rate of R [bits/sec]. The message transmission time is assumed to be exponentially distributed. The average message length is equal to 1500
[bits/message]. Messages arrive at a rate of 0.5 [messages/sec].
Design the system by choosing the lowest value of the channel transmission rate
R assuring that all the following objectives are met:
The average message waiting time is lower than or equal to twice its average
service time.
The probability that the message waiting time is longer than 3 times its mean
service time is not higher than 0.01 (i.e., its 99percentile waiting time value
is not higher than 3 times its mean service time).
Hint: Use Matlab to solve an equation. For example, if you would like to
solve x2 + 4x + 1 = 0, the corresponding Matlab code is give as
syms x
solve(x2 + 4 x + 1 == 0)
Ans:
The arrival rate is 0.5 [message/sec], the service rate is
= 0.5
= 750
.
R
R

R
1500

[message/sec] and, thus,

1500

The average waiting time is


R 1125 [message/sec].

(1)

2 . We obtain that =

750
R

2
3

and, thus,

P (W > 3 ) = e(1) 0.01. We obtain 0.1343 and, thus, R 5.5845K


[message/sec].
Therefore, we conclude that R 5.5845K [message/sec].

Problem 3
Consider a concentrator that queues messages generated by local terminals and
transmits them across a single outgoing communications trunk. Messages arrive in
accordance with a Poisson process at a rate of one message every 0.004 seconds. The
message transmission times modeled to be exponentially distributed with a mean of
0.0025 seconds. Assume steady state conditions.
(a) Calculate the mean number of messages in the system.
(b) Calculate the mean message waiting time.
(c) Suppose we do a survey which finds that customers are willing to continue to use
this system as long as the mean message waiting time does not increase to more
than twice the value calculated in part b. Under such a targeted performance
level, to what value can the message arrival rate be increased?
2

EE132B-HW Set #9

Ans:
Arrival rate is
= 58 .

1
0.004

UCLA 2014 Fall

= 250 [message/sec], service rate is

Prof. Izhak Rubin

1
0.0025

= 400 [message/sec] and

(a)
The mean system size is E[X] =

5
3

[message].

(b)
The mean waiting time is E[W ] =

(1)

= 0.0042 [sec].

(c)
We denote the increased arrival rate as 0 and 0 as
E[W 0 ] =
We obtain 0

2
1+

0
.

We have:

0
0

= 2 0.0042.
(1 0 )

(1)

= 307.6923 [message/sec].

Problem 4
Consider a transmission node that queues messages generated by local terminals
and transmits them across two outgoing communications channels. Messages arrive
in accordance with a Poisson process at a rate of one message every 0.1 seconds. The
message transmission times modeled to be exponentially distributed with a mean of
0.16 seconds. Assume steady state conditions.
(a) Calculate the mean number of messages in the system.
Hint: the mean system size for an M/M/2 queueing system is
"

f2
P (0) f +
2!
where f =

and =

2
+
(1 )2 1

!#

(2)

.
2

(b) Calculate the mean message waiting time.


(c) Assume now that the system provides no waiting (buffering) facility. Thus,
when the number of messages in the system is equal to 2, no new arrivals are
admitted; new arrivals are then blocked. Calculate the probability that an
arriving message is blocked.
Ans:
Arrival rate =
f = = 1.6; =

1
0.1

= 10 [message/sec]; service rate =


= 0.8, where m = 2.
3

1
0.16

= 6.25 [message/sec];

EE132B-HW Set #9

UCLA 2014 Fall


fj
,
j!

We have a0 = 1, aj =

fj
,
m!mjm

for j < m, aj =

Prof. Izhak Rubin


for j m. Then, P (0) = P1
j=1

aj

and P (j) = aj P (0), for j = 1, 2, . . . , . Therefore, we conclude

m1
X

P (0) =

j=0

fj fm 1
+
f
j!
m! 1 m

"

f2 1
= 1+f +
2 1

#1

P (j) = aj P (0), j 1.
(a)
The mean number of messages in the system is given as
E[X] =

jP (j) = P (0)

j=1

jaj

j=1

f jm
fm X
j jm
= P (0) f +
m! j=m m

fm X
= P (0) f +
(j + m)m
m! j=0

fm
= P (0) f +
m!
= 4.4444.
"

m
+
2
(1 )
1

!#

(b)
The mean system delay is E[D] = E[X]
= 0.4444 [sec]. Then, the mean waiting

1
time is E[W ] = E[D] = 0.2844 [sec].
(c)
Now, system only allows two messages. Hence, we have
P (0) =

1
1+f +

f2
2

= 0.2577,

P (j) = aj P (0), j = 1, 2.
The blocking probability PB = P (2) = 0.3299.

Potrebbero piacerti anche