Assignment #2

2102522 Internet Technology and Applications

=======================================================================

1. a) Find P[ N  n] for M/M/1 system.

b) What is the maximum allowable arrival rate in a system with service rate  if we require
that P[ N  10] = 10−3

2. Consider an M/M/1 queueing system with arrival rate  customers/second. (10 points)
a) Find the service rate required so that the average queue is five customers (i.e., E[ N q ] = 5 ).
b) Fid the service rate required so that the queue that forms from time to time has mean 5 (i.e.,
E[ N q | N q  0] = 5 ).
c) Which of the two criteria, E[ N q ] = 5 or E[ N q | N q  0] = 5 , do you consider the more
appropriate.

3. Consider an M/M/1 queueing system in which each customer arrival brings in a profit of $5
but in which each unit time of delay costs the system $1. Find the range of arrival rate for which
the system makes a net profit.

4. For an M/M/1/ queueing model, assume that the arrival rates of new customers decreases
with the number of customers in the system as follows:

k = โดย k = 0,1, 2,
k +1
while the departure rate  remains a constant value. (10 points)
a) draw the state transition diagram of this queueing model.
b) Find the state probabilities
c) Find the average number of customers in the system
d) Find the effective arrival rate
e) Find the mean time each customer spent in the system

5. A decision to purchase one of two machines is to be made. Machine 1 has a processing rate of
 transactions/hour and it costs B dollars/hours to operate; Machine 2 is twice as fast but costs
twice as much to operate. Suppose that transactions arrive at the system according to a Poisson
process of rate  and that the transaction processing times are exponentially distributed. The
total cost of the system is the operation cost plus a cost of A dollars for each hour a customer
has to wait.

a. Find expressions for the total cost per hour for each of the systems. Plot this cost versus the
arrival rate.
b. If A = B / 10 , for what range of the arrival rates is machine 1 cheaper? Repeat for A = 10 B

6. Consider an M/M/1/2 queueing system in which each customer accepted into the system
brings in a profit of $5 and each customer rejected results in a loss of $1. Find the arrival rate at
which the system breaks even.

7. Consider an M/M/5/5 system in which the arrival rate is 10 customers per minute and the
mean service time is 1/2 minute.

a. Find the probability of blocking a customer. Hint: Use the recursive formula of Erlang B.

b. How many more servers are required to reduce the blocking probability to 10%?

8. For the Erlang B formula which is denoted by

ac
B ( c, a ) = c!
 a 2
a3 ac 
1 + a + + +  + 
 2! 3! c! 
where B(c, a) is the blocking probability
a the traffic intensity and a =  / 
c is the number of servers
show that the Erlang B formula satisfies the following recursive equation:
aB(c − 1, a)
B ( c, a ) =
c + aB(c − 1, a)

9. For the Erlang C formula,

pc
C ( c, a ) = P[W  0] = P[ N  c ] =
1− 
where
ac
 
pc = c −1 k c !c  , a = and  =
 c
 
a a
+  k −c
k =0 k ! c ! k =c
show that
cB(c, a)
C (c, a ) = for c  a
c − a (1 − B (c, a) )

10. Suppose that department A in a certain company has three private telephone lines
connecting two sites. Calls arrive according to a Poisson process of rate 1 call/minute, and have
an exponentially distributed holding time of 2 minutes. Calls that arrive when three lines are
busy are automatically redirected to public telephone lines. Suppose that department B also has
three private telephone lines connecting the same sites, and that it has the same arrival and
service statistics.
a. Find the proportion of calls that are redirected to public lines.
b. Suppose we consolidate the telephone traffic from the two departments and allow all calls to
share the six lines. What proportion of calls is redirected to public lines?

11. Customers arrive at a shop according to a Poisson process of rate 12 customers per hour.
The shop has two clerks to attend to the customers. Suppose that it takes a clerk an
exponentially distributed amount of time with mean 5 minutes to service one customer.
a. What is the probability that an arriving customer must wait to be served?
b. Find the mean number of customers in the system and the mean time spent in the system.
c. Find the probability that there are more than 4 customers in the system.

12. Find P[ N  c + k ] for an M/M/c system.

13. Little’s formula applied to the servers implies that the mean number of busy servers is
 E[ ] . Verify this by explicit calculation of the mean number of busy servers in an M/M/c
system.

14. A tool rental shop has four floor sanders. Customers for floor sanders arrive according to a
Poisson process at a rate of one customer every two days. The average rental time is
exponentially distributed with mean two days. If the shop has no floor sanders available, the
customers go to the shop across the street.

a. Find the proportion of customers that go to the shop across the street.

b. What is the mean number of floor sanders rented out?

c. What is the increase in lost customers if one of the sanders breaks down and is not replaced?

c. Find E[W ] and E[T ] .

d. For c = 2, compare E[W ] and E[T ] for this system to those of M/M/1 and M/M/2 systems of
the same maximum processing rate.

15. a. show that the Erlang C formula is related to the Erlang B formula by

cB(c, a)
C (c, a ) = for c  a
c − a[1 − B(c, a )]

b. Show that this implies that C (c, a)  B(c, a) .

(  )
 k

16. Show that B(c, a)  k =c k!

e ( )  C (c, a )
−  
17. Consider a queueing system in which the maximum processing rate is c customers per
second. Let k be the number of customers in the system. When k  c, c customers are served
at a rate  each. When 0  k  c, these k customers are served at a rate c / k each. Assume
Poisson arrivals of rate  and exponentially distributed times.
a. Find the transition rate diagram for this system.
b. Find the steady state pmf for the number in the system.

18. Suppose that department A in a certain company has three private telephone lines
connecting two sites. Calls arrive according to a Poisson process of rate 1 call/minute, and have
an expontentially distributed holding time of 2 minutes. Calls that arrive when three lines are
busy are automatically redirected to public telephone lines. Suppose that department B also has
three private telephone lines connecting the same sites, and that it has the same arrival and
service statistics.

a. Find the proportion of calls that are redirected to public lines.

b. Suppose we consolidate the telephone traffic from the two departments and allow all calls to
share the six lines. What proportion of calls are redirected to public lines?

19. In the BTS system near Siam Square, there are two public phone booths available in a small
corner where at most two customers may be queued in case both
phone booths are occupied. The average calling time is exponentially distributed with mean one
minute. Customers do not wish to wait in the queue for long, such that the arrival rates of
customers are 0 = 1.2, 1 = 1.2, 2 = 0.8 and 3 = 0.4 customers/minute. (10 points)

a) Draw the system state transition diagram.

b) Find the global balance equations in the given Table.

c) Find the state probabilities

d) What is the blocking probability?

e) Find the average number of customers in the systems.

f) The average time each customer spends in the system

20. Suppose we use P[ N = c] from an M/M/ system to approximate B(c, a) in selecting the
number of servers in an M/M/c/c system. Is the resulting design optimistic or pessimistic?

21. A computer is shared by 15 users. Suppose that the mean service time is 2 seconds and the
mean think time is 30 seconds, and that both of these times are exponentially distributed.

a. Find the mean delay and mean throughput of the system.

b. Repeat part a if 5 users are added to the system.

22. Find the transition rate diagram and steady state pmf for a two-server finite-source
queueing system.

23. Consider a c-server, finite-source queueing system that allows no queueing for service.
Requests that arrive when all servers are busy are turned away, and the corresponding source
immediately returns to the “think” state, an spends another exponentially distributed think time
before submitting another request for service.

a. Find the transition rate diagram and show that the steady state pmf for the state of the system
is

K j k− j
  p (1 − p )
PK [ N − j ] = C 
j
i = 0,1, ,c
K j
   p (1 − p )
i =0 
k− j

j 

where c is the number of servers, K is the number of sources, and

/
p= .
1+ / 

b. Find the probability that all servers are busy.

c. Use the fact that arriving customers “see” the steady state pmf of a system with one less
source to show that the fraction of arrivals that are turned away is given by PK −1 (c) . The
resulting expression is called Engset formula.