QUEUEING SYSTEMS

• NOMENCLATURE OF A QUEUEING SYSTEM

• The origin of queuing theory can be traced back to A.K. Erlang, a Danish
engineer who applied queuing theory extensively to the study of the behaviour
of telephone networks.
• Erlang developed several results that are still used in queuing performance
evaluation today.
• Basically, a queuing system can be broken down into three major components:
the input process, the system server, and the output process.
• If an arriving customer finds the service facility occupied (all servers busy), he
joins the waiting queue; and he therefore receives his service later on when he
reaches the head of the waiting queue.
• In the case of data switching network, customers are packets or messages that
arrive at a switching node, and servers are the transmission channels.
• In a CPU job scheduling problem, customers are computer processes, and
servers are the various computer resources such as CPU, I/O devices, etc.

Elementary Queueing Capacity Analysis 1

Theory
 QUEUEING SYSTEMS
• NOMENCLATURE OF A QUEUEING SYSTEM - INPUT PROCESS
• When we talk about the input process, we are concerned about three aspects:
• 1. Size of the arriving population: The number of arriving customers have a
direct impact on queuing results. Telephone customers requesting telephone
lines and air travelers calling an air ticket reservation system are examples of
infinite calling population. However, they are finite but large populations.
• 2. Arriving patterns: Customers may arrive at a queuing system either in
some regular patterns or at random. The most commonly assumed arrival
process is the Poisson process. Some probability distributions normally used to
describe arrival processes are: M – Markovian or Poisson process; D-
Deterministic, fixed inter-arrival times; EK – Erlang distribution of order k;
and G – General probability distribution.
• 3. Behaviour of Arriving Customers: Customers arriving at a queue may
behave differently when the system is full (due to a finite queue) or when all
servers are busy. If an arriving customer leaves and is considered lost without
entering the system when the system is full, that system is called a blocking
system.

Elementary Queueing Capacity Analysis 2

Theory
 QUEUEING SYSTEMS
• QUEUEING SYSTEM STRUCTURE

Arriving Departing
customers customers

• A service facility may comprise one or more servers.

• In a system with parallel and identical servers, a customer at the head of a
waiting queue can go to any server that is free, and leave the system after
receiving service.
• The system capacity, C, refers to the maximum number of customers that a
queuing system can accommodate.
• In a multiserver queuing system, the system capacity is the sum of the
maximum size of the waiting queue and the number of servers.

Elementary Queueing Capacity Analysis 3

Theory
 QUEUEING SYSTEMS
• CHARACTERISTICS OF THE OUTPUT PROCESS
• The queuing discipline refers to the manner in which customers in a waiting
queue are selected and served. In general we have:
• A) First-come-first-serve (FCFS) – the most frequently encountered.
• B) Last-come-first-serve (LCFS)
• C) Priority service
• D) Processor sharing

• Typical service distributions are: M – Markovian or Poisson process; D-

Deterministic, fixed inter-arrival times; EK – Erlang distribution of order k;
and G – General probability distribution.
• The most commonly assumed service time distribution is the exponential
distribution.

Elementary Queueing Capacity Analysis 4

Theory
 QUEUEING SYSTEMS
• KENDALL NOTATION
• The Kendall notation for a queuing system is:

A/ B / X /Y / Z

• Where:
A: customer arrival pattern (inter-`arrival time distribution)
• B: Service pattern (service time distribution)
• X: Number of parallel servers
• Y: System capacity
• Z: Queuing discipline
• For example, M/M/1//FCFS represents a queuing system whose customers
arrive according to a Poisson process and receive exponentially distributed
service times. The system has only one server, an infinite waiting queue, and
customers are served on a first-come-first-serve basis.
• In many cases, we only use the first three parameters, with the default values
for the last two parameters being Y= and Z=FCFS.

Elementary Queueing Capacity Analysis 5

Theory
 QUEUEING SYSTEMS
• LITTLE’S THEOREM
• Little’s theorem states that, for a average number of customers (N) in a steady-
state queuing system, with an average arrival rate l of customers entering the
system, and the average time, T, a customer spends in the system, we have:
• N  lT
• This result does not assume any specific distribution for the arrival as well as
the service process, nor does it assume any queuing discipline or depend upon
the number of parallel servers in the system.
• The time average value for a stochastic process, X(t), is obtained by
observing the process over a long time, T, and then taking the average as:
1T
m  lim  X (t )dt
t  T 0

• In probability theory, the expected value of a process at time T refers to the

average value of various realizations at time T:

E  X (T )   kP X (T )  k 
k 0

• For a stochastic process, if the time average is equal to the ensemble average,
we say the process is ergodic.

Elementary Queueing Capacity Analysis 6

Theory
 QUEUEING SYSTEMS
• RESOURCE UTILIZATION & TRAFFIC INTENSITY
• Resource utilization represents the fraction of time a server is engaged in
providing service.
Time a server is occupied
Utilizatio n( r ) 
Time server is available
• For a queuing system with m (m1) servers, if there are N customers in a time
interval (t, t+T), then each server, on average, will serve (lT/m) customers.
• If the average service time =1/m, then the utilization, r, becomes:
r
lT / m 1 / m   l
T mm
• It is clear that r should be dimensionless and less than unity (r<1) in order for
a server to cope with the service demand.
• The traffic intensity, or offered load, measured in Erlangs, is the product of
the arrival rate (l), and the average service time, (1/m):
l
Traffic int ensity a 
m

Elementary Queueing Capacity Analysis 7

Theory
 QUEUEING SYSTEMS
• EXAMPLES

• 1. Jobs arrive at a single processor system according to a Poisson

process with an average of 10 jobs per second. Determine:
• a) The probability that no job arrives in a 0ne-second period.
• b) The probability that five or fewer jobs arrive in a one-second period.

• 2. A Private Branch Exchange (PBX) was designed to handle the voice

traffic generated by 300 employees in a factory. If each employee, on
average, makes 1.5 calls per hour, with an average call duration of 2.5
minutes, what is the offered load presented to this PBX?

Elementary Queueing Capacity Analysis 8

Theory
 QUEUEING SYSTEMS
• THE POISSON PROCESS
• The Poisson process is central to the physical process modeling and plays a
pivotal part in classical queuing theory.
• In most of the elementary queuing systems, the inter-arrival times and the
service times are assumed to be exponentially distributed, or, equivalently,
the arrival and service processes are assumed to be Poisson.
• The Poisson process is a counting process: the random variable X(t) that
counts the number of point events in a given time interval (0,t) has a
Poisson distribution if:

P X (t )  k  
lt k  lt
e
k!

Point events
Poisson point events

Time
0 t

Elementary Queueing Capacity Analysis 9

Theory
 QUEUEING SYSTEMS
• THE POISSON PROCESS – SUPERPOSITION PROPERTY
• In the foregoing formula, (lt) is the mean of a Poisson variable: physically it
represents the average number of occurrences of the event in time interval
t, hence l can be interpreted as a rate.
• The superposition property states that if N independent Poisson processes
A1,A2,..AN, are combined into a single process, A=A1+A2+A3+..+AN, then
process A is still Poisson with rate l equal to the sum of the individual rates
li of Ai.

Poisson A1

+ Poisson A

Poisson AN

Elementary Queueing Capacity Analysis 10

Theory
 QUEUEING SYSTEMS
• THE POISSON PROCESS – SUPERPOSITION PROPERTY
• To prove this, recall that the generating function, P(z), for a Poisson
distribution is:
P( z )  e  lt (1 z )

• Since A=A1+A2+A3+..+AN, we have:

  
E z A  E z A  z A  z A  ..  z A
1 2 3 N

 E z E z .....E z 
A1 A2 AN

 e  l t (1 z ) .....e  l
1 N
t (1 z )

 e  l  l1 2  ..  l N t (1 z )
 e  lt (1 z )

• The right-hand-side is the transform of a Poisson distribution with rate

l  l1  l2  ..  lN

• Hence the resultant process is Poisson.

Elementary Queueing Capacity Analysis 11

Theory
 QUEUEING SYSTEMS
• EXPONENTIALLY DISTRIBUTED INTERARRIVAL TIMES
• The exponential distribution and the Poisson distribution are closely related,
and mirror each other in the following sense:
• - If the inter-arrival times in a point process are exponentially distributed,
then the number of arrival points in the time interval (0,t) is given by a Poisson
distribution, and the process is a Poisson arrival process.
• - Conversely, if the number of arrival points k, in any interval (0,t), is a
Poisson random variable, the inter-arrival times are exponentially distributed,
and the arrival process is Poisson.
• Let t be the inter-arrival time. Then:
Pt  t   1  Pt  t 
• But P[t>t] is the probability that no arrivals occur in the interval [0,t]. That is:

Pt  t   P0 (t )  e  lt

• Therefore we obtain the exponentially distributed inter-arrival time:

Pt  t   1  e  lt

Elementary Queueing Capacity Analysis 12

Theory
 QUEUEING SYSTEMS
• THE POISSON PROCESS – MEMORYLESS PROPERTY
• The memoryless property of a Poisson process means that if we observe at a
certain point in time, the distribution of the time until the next arrival is not
affected by the fact that some time has passed since the last arrival.
• In other words, the process starts afresh at the time of observation and has no
memory of the past.
• The memoryless property states that the distribution of the remaining time
until the next arrival, given that to units of time have elapsed since the last
arrival, is identically equal to the unconditional distribution of inter-arrival
times.
• Assume that we start observing the process immediately after an arrival at time
t=0. From the Poisson distribution, we see that the probability of no arrivals
in the time interval (0,to) is given by:


P 0 arrivals in 0, t0   lt0 0  lt
e o
 e  lt o

0!

Elementary Queueing Capacity Analysis 13

Theory
 QUEUEING SYSTEMS
• THE POISSON PROCESS – MEMORYLESS PROPERTY
• We now determine the conditional probability that the first arrival occurs in
(to, to+t), given that a period to has elapsed. That is:


P arrival in t0 , t0  t  / no arrivals in 0, t0  
t0 t
 lt
 le dt

t0


 e  l t  t   e  lt
0 0

e  lt 0
e  lt 0

 1  e  lt
• But the probability of an arrival in (0,t) is also given by:
 
P arrival in (0, t )  1  P0  1  e  lt

• Therefore we see that the conditional probability of inter-arrival tines, given

that a certain time has elapsed, is the same as the unconditional distribution.
This confirms the Markovian or memoryless property of the Poisson counting
process and the exponential distribution.

Elementary Queueing Capacity Analysis 14

Theory
 QUEUEING SYSTEMS
• Assignment 6:
• 1. A communications channel operating at 9600 bps receives two types of
packet streams from a gateway. Type-A packets have a fixed length format of
48 bits, whereas Type-B packets have an exponentially distributed length with
a mean, l, of 480 bits. If, on the average, there are 20% Type-A packets and
80% Type-B packets, calculate the utilization of the channel assuming the
combined arrival rate is 15 packets per second.

• 2. All telephone lines in Adama Science and Tecnology University are

connected to the University central switchboard which has 120 external lines
to the local telephone exchange. The voice traffic generated by its employees
in a typical working day is as follows:
• Local Calls: 500 calls/hr (incoming); 480 calls/hr outgoing; 2.5 min mean holding time
• Trunk calls: 30 calls/hr incoming; 10 calls/hr outgoing; 1 min mean holding time.
• Find a) The total traffic offered to the PABX; b) The overall mean holding time of the
incoming traffics; c) The overall mean holding time of the outgoing traffic.

Elementary Queueing Capacity Analysis 15

Theory
 QUEUEING SYSTEMS
• BIRTH-DEATH PROCESS
• A birth-death process is a special case of a Markov chain in which the process
makes transition only to the neighboring states of its current positions.
• This restriction simplifies treatment and makes it an excellent model for all
basic queuing systems – the so-called elementary queuing theory.
• The term birth-death implies that when the population is k at time t, the
process is said to be in state Ek at time t. A transition from Ek to Ek+1 signifies
a birth, and a transition down to Ek-1 signifies a death in the population.
• Implicitly, there can be no simultaneous births or deaths in an incremental
time period Dt. The state-transition diagram is shown below:

k-1 k k+1

Elementary Queueing Capacity Analysis 16

Theory
 QUEUEING SYSTEMS
• BIRTH-DEATH PROCESS
• If we define lk to be the birth rate when the population size is k, and mk the
death rate when the population size is k, then we have:
lk  qk , k 1; m k  qk , k 1
• Substituting these into the transition-rate matrix, [Q], we have:
 l0 l0 0 0 
 m  l1  m1  l1 0 0
 1 
Q   0 m2  l2  m 2  l2 0
 
 0 0 m3  l3  m 3  l3 
 0 0 0  

where q jj    q ji  q00   q01  l0 ; q11   q10  q12  ( m1  l1 );

i j

q22   q21  q23  m 2  l2 ; q33   q32  q34  m3  l3 ; etc.

• Expanding and using the more familiar notation, we have:

d
Pk (t )  lk  m k Pk (t )  lk 1 Pk 1 (t )  m k 1 Pk 1 (t ), k  1
dt
d
P0 (t )  l0 P0 (t )  m1 P1 (t )
dt
Elementary Queueing Capacity Analysis 17
Theory
 QUEUEING SYSTEMS
• BIRTH-DEATH PROCESS
• A special case of the birth-death process is the pure-birth process, where:

lk  l k  0,1,2,..
m k  0, k  0,1,2,..

• Assuming the initial condition is Po(0)=0, we have, from the above two
equations:

d
Pk (t )  lPk (t )  lPk 1 (t ) k  1
dt
d
P0 (t )  lP0 (t )
dt

• The solution is:

P (t ) 
lt k  lt
e
k
k!

Elementary Queueing Capacity Analysis 18

Theory
 QUEUEING SYSTEMS
• CLASSICAL M/M/1 QUEUING SYSTEMS
• A Markovian queuing system is one which is characterized by a Poisson
arrival process and exponentially distributed service times.
• First, we compute the probability, Pk, of having k customers in the queuing
system under equilibrium conditions, then calculate N from the expression:

N   kPk
k

• Once N is found, we can then evaluate other parameters of interest to us,

using Little’s theorem.

l m

The M/M/1 Queue

Elementary Queueing Capacity Analysis 19

Theory
 QUEUEING SYSTEMS
• CLASSICAL M/M/1 QUEUING SYSTEMS
• The classical M/M/1 queue refers to a queuing system where customers arrive
according to a Poisson process (memoryless process), and are served by a
single server with an exponential service time (memoryless).
• The arrival rate l and the service rate, m, do not depend upon the number of
customers in the system; in other words, it is state-independent.
• In the Kendall notation, capacity is assumed to be infinite, and customers are
served on a FCFS basis.
• Let N(t) be the number of customers in the system in time t. A customer
arriving in the system is considered a birth; and a customer leaving the
system is considered a death.
• Since the Poisson process rules out the possibility of having more than one
arrival in Dt, and the exponential service time ensures that there is, at most,
one departure in Dt, N(t) is a birth-death process because it can only go to
the neighbouring states – N(t)+1 or N(t)-1, in an incremental time Dt.

Elementary Queueing Capacity Analysis 20

Theory
 QUEUEING SYSTEMS
• CLASSICAL M/M/1 QUEUING SYSTEMS
• From the earlier expressions, we use the equations governing a birth-death
process:
d
Pk (t )  lk  m k Pk (t )  lk 1 Pk 1 (t )  m k 1 Pk 1 (t ), k  1
dt
d
P0 (t )  l0 P0 (t )  m1 P1 (t )
dt
• Generally, in network performance evaluation, we are only interested in the
long-term (steady-state) behaviour, at which:
d
Pk (t )  0  Pk  lim Pk (t )
dt t 

• Since the birth rates and death rates are state-independent, we have:
lk  l ; m k  m
• Then the equations (known as GLOBAL BALANCE EQUATIONS) become:
l  m Pk  lPk 1  mPk 1
mP1  lP0

Elementary Queueing Capacity Analysis 21

Theory
 QUEUEING SYSTEMS
• The same comments apply regarding the conservation of flow across any
boundary.
• Rather than surrounding each state and writing down its equations (the net
flow into state k must equal the net flow out of state k), we could draw an
artificial line between states k and k+1. In this case, the mean flows across the
barrier must be equal to zero under steady-state conditions.
• That is, the flow to the right must equal the flow to the left. This yields the
detailed balance equation:

lPk  mPk 1
• If we consider a barrier between states k-1 and k, we get the detailed balance
equations as:

lPk 1  mPk
• Sometimes it may be advantageous to work with detailed rather than global
(state-based) balance equations for the steady-state probabilities.

Elementary Queueing Capacity Analysis 22

Theory
 QUEUEING SYSTEMS
• CLASSICAL M/M/1 QUEUING SYSTEMS
• The first equation is a recursive equation which may be solved using the z-
transform:   
 l  m Pk z   lPk 1 z   mPk 1 z k
k k

k 1 k 1 k 1

l  m P( z )  P0   lzP ( z )  m P( z )  P1 z  P0 

z
• Coupling with the boundary condition, where (P1=(l/m)Po, we have:
P0 m 1  z  P0 m 1  z 
P( z )  
lz 2  l  m z  m lz  m  z  1
P0
P( z ) 
1  l / m z
• To evaluate Pk, we set:
 Pk  1  P( z ) z 1  1
k
• This gives:
P0  l
 1  P0  1  
1 l / m  m

Elementary Queueing Capacity Analysis 23

Theory
 QUEUEING SYSTEMS
• CLASSICAL M/M/1 QUEUING SYSTEMS
• Setting r  l / m  P0  1  r
• Then we obtain:
1 r
P( z ) 
1  rz
Pk  1  r r k

• Therefore the steady-state probability of having k customers in the system is a

geometric random variable with parameter r.
• For stability, we require that r<1.
• The probability of having n or more customers in the system is given by:
   n 1
PN  n   Pk  1  r   r  1  r   r   r k 
k  k

k n k n k  0 k 0 
 1 1 r n 
PN  n  1  r     rn
1  r 1  r 
 P N  n   r n

Elementary Queueing Capacity Analysis 24

Theory
 QUEUEING SYSTEMS
• CLASSICAL M/M/1 QUEUING SYSTEMS
• The average number of customers in the system, N, is given by:
  
N   kPk   k 1  r r r 1  r   kr k 1
k

k 0 k 0 k 0

r l
N  
1 r m l

In practice, for stability, the design rule is r<2/3.

Elementary Queueing Capacity Analysis 25
Theory
 QUEUEING SYSTEMS
• CLASSICAL M/M/1 QUEUING SYSTEMS
• Average time, T, a customer spends in the system, or the system delay, is:
Average No. of Customers N r 1
T   
Arrival rate l l 1  r  m  l 
• The average number of customers at the service facility:
Ns  l / m  r
• Thus the number of customers being served is equal to the offered load, or the
utilization.
• The average time spent in the waiting queue, or waiting time, W (excluding
service time) is:
1 r
W T  
m m  l 
• Finally, the average number of customers in the waiting queue is given by:
r2
N q  lW 
1  r 

Elementary Queueing Capacity Analysis 26

Theory
 QUEUEING SYSTEMS
• M/M/1/S QUEUING SYSTEMS
• In a realistic M/M/1 queuing system, only S customers can be
accommodated. The state transition-rate diagram is the same as the classical
M/M/1, except that it is truncated at state S, as shown below:

l l l l

0 k-1 k k+1 S
m m m
m
The M/M/1/S Transition-rate diagram
• The truncation will affect queuing results. Using the normalization equation,
but summing only to S, we have:
S 1 r
P0  r k  1  P0 
k 0 1  r S 1

 Pk 
1  r r k
1  r S 1

Elementary Queueing Capacity Analysis 27

Theory
 QUEUEING SYSTEMS
• M/M/1/S QUEUING SYSTEMS – PROBABILITY OF BLOCKING
• The probability of blocking is the probability that customers are blocked and
not accepted by the queuing system because the system capacity is exceeded.
• When there are S customers in the system, we say the system is saturated, and
Ps  PN  S  
1  r r S
1  r S 1
• Then the probability of blocking is the probability of saturation:

Pb  Ps 
1  r r S
1  r S 1
• Average number of customers in the system:

S S  1 r   1 r  S
N   kPk    
S 1 
kr k
   r kr k 1
S 1  
k 0 k  0 1  r   1  r  k 0
r r
N   S  1Ps
1 r 1 r

Elementary Queueing Capacity Analysis 28

Theory
 QUEUEING SYSTEMS
• M/M/1/S QUEUING SYSTEMS
• Average number of customers at the service facility is:
N s  Pk  0E N s / k  0  Pk  0E N s / k  0
N s  1  P0  r 1  Ps 
• Average number of customers at the waiting queue is:
r2 S  r
Nq  N  NS   Ps
1 r 1 r
• Average time spent in the system (T) and in the waiting queue (W). Since
customers are blocked when S customers are in the system, the effective
arrival rate of customers admitted into the system is:
l '  l 1  Ps 
• Then we find that:
• N 1 Sr S 1
T  
l ' m  l l  mr S 1
Nq Sr S 1
r
W  
l ' m  l l  mr S 1
Elementary Queueing Capacity Analysis 29
Theory
 QUEUEING SYSTEMS
• MULTI-SERVER SYSTEMS – M/M/m
• Here, the service facility consists of m identical parallel servers.

l
m servers

A multi-server queuing system

• The M/M/m transition-rate diagram is shown below:

l l l l l

0 1 2 m k
m 2m mm mm mm
M/M/m transition-rates diagram

Elementary Queueing Capacity Analysis 30

Theory
 QUEUEING SYSTEMS
• MULTI-SERVER SYSTEMS – M/M/m
• Let us focus our attention on the system state, N(t) that is a birth-death process
with state-dependent service rates.
• When there is one customer, one server provides service at rate m; if there are
two customers, then two servers are engaged, and the service rate is 2m; and
the service rate increases until mm and stays constant thereafter.
• Let us assume an infinite buffer; then for local balance, traffic leaving state k
to state k+1 must equal traffic leaving state k+1 to state k:
k  m : kmPk  lPk 1
k  m : mmPk  lPk 1
• Recalling that utilization r and traffic intensity (a) are defined as:
l l
r ; a
• For k<m, we have: mm m

l  l  l 
Pk  Pk 1     Pk  2
km km
  ( k  1) m 
ak
 Pk  P0 , k  m
k!
Elementary Queueing Capacity Analysis 31
Theory
 QUEUEING SYSTEMS
• MULTI-SERVER SYSTEMS – M/M/m
• For km, we have:
k m k m
l l 1 ak m  am 
Pk  Pk 1      Pm  k  m  P0 
mm m m m  m! 
ak
Pk  P0
m!m k  m
• Using the normalization condition, where sum of Pk=1, we get:

m 1a k  ak 1 
P0     k m 
1
 k  0 k ! k  m m! m 
1
 P0 
m 1a k am 
  k!  m!1  r 
k  0 

Elementary Queueing Capacity Analysis 32

Theory
 QUEUEING SYSTEMS
• MULTI-SERVER SYSTEMS – M/M/m
• The probability that an arriving customer will find all servers busy and will be
forced to wait in the queue is:
P0 mp   k  m
m
 P0 a k

Pd  Pdelay   Pk   k m
 r
k m k  m m! m m! k  m
P0 a m
 Pd 
m!(1  r )

• This is the Erlang-C formula, or the Erlang delay formula, or Erlang

formula of the second kind.
• Number of customers waiting in the queue:
  P0 a m  k am  k am r
N q   kPm  k   P  kr  P
m! 1  r 2
k 0 0
k 0 k  0 m! m m! k  0
r l
 Nq  Pd  Pd
1 r mm  l

Elementary Queueing Capacity Analysis 33

Theory
 QUEUEING SYSTEMS
• MULTI-SERVER SYSTEMS – M/M/m
• The time spent in the waiting queue is given by:
Nq Pd
W 
l mm  l

• The time spent in the queuing system is:

1 Nq Pd 1
T W    
m l mm  l m

• The number of customers in the queuing system is:

r
N  lT  Pd  mr
1 r

Elementary Queueing Capacity Analysis 34

Theory
 QUEUEING SYSTEMS
• EARLANG’S LOSS LOSS QUEUING SYSTEMS: M/M/m/m QUEUE
• Here, we are looking at a realistic multi-server system with a finite waiting
queue.
• The state transition rates diagram is the same as that of the M/M/m queue
except it is truncated at state m (blocked at m).
• Using the local balance concept:
lPk 1  kmPk
 ak 
Pk    P0
 k! 
Solving for Po, we obtain:

P0 
1
 Pk 
 
a k / k!
 m ak   m ak 
  k!    k! 
 k  0  k  0 

Elementary Queueing Capacity Analysis 35

Theory
 QUEUEING SYSTEMS
• EARLANG’S LOSS LOSS QUEUING SYSTEMS: M/M/m/m QUEUE
• The probability that an arrival will be lost when all servers are busy is given
by:
Pb 
a
m
/ m! 
 m ak 
  k! 
 k 0 

• This is the blocking probability, also called Erlang B formula or Erlang’s

loss formula, denoted as B(a,m).
• Number of customers in the queuing system is:
 m 1a k
N   kPk  P0   a (1  Pb )
k 0 k 0 k!
• Since there is no waiting, the system delay is the same as the service time:

1
T
m

Elementary Queueing Capacity Analysis 36

Theory
 QUEUEING SYSTEMS
• FINITE CUSTOMER POPULATION, SINGLE
SERVER: M/M/1//M QUEUE
• Here we consider the case where we no longer have a Poisson input
process with infinite population, but rather a finite population of
possible users.
• The system structure is such that we have a total of M users; a
customer is either in the system (consisting of a queue and a single
server) or outside the system and in some sense “arriving”.
• In particular, when the customer is in the “arriving” condition, then the
time it takes him to arrive is a random variable with an exponential
distribution whose mean is 1/l seconds.
• All customers are assumed to act independently of each other. As a
result, where there are k customers in the system (queue plus service),
then there are (M-k) customers in the “arriving” state and therefore the
total arrival rate in this state is l(M-k).

Elementary Queueing Capacity Analysis 37

Theory
 QUEUEING SYSTEMS
• FINITE CUSTOMER POPULATION, SINGLE
SERVER: M/M/1//M QUEUE
• We therefore see that this system is in a strong sense self-regulating in
the sense that when the system gets busy, then the rate at which
additional customers arrive is in fact reduced, thus lowering the further
congestion of the system.
• We model this system quite appropriately with out birth-death process
by choosing:

l ( M  k ) 0  k  M
lk  
 0 kM
m k  m , k  1,2,3,..

Elementary Queueing Capacity Analysis 38

Theory
 QUEUEING SYSTEMS
• FINITE CUSTOMER POPULATION,
SINGLE SERVER: M/M/1//M QUEUE
• We assume that we have sufficient room to contain M customers in the
system.
• The finite state-transition-rate diagram is shown below.

Ml (M-1)l 2l l

0 1 2 M-2 M-1 M

m m m m

State-transition-rate diagram for single-server finite-population

system, M/M/1//M

Elementary Queueing Capacity Analysis 39

Theory
 QUEUEING SYSTEMS
• FINITE CUSTOMER POPULATION, SINGLE
SERVER: M/M/1//M QUEUE
• Using the expressions for a birth-death process, we have the equilibrium
probabilities pk in terms of a single unknown constant po:
k 1
li k 1
l M  i 
p k  po   po  , 0k M
i 0 m i 1 i 0 m
  l k M !

 pk   po  m  M  k ! 0  k  M
 
 kM
 0
• To determine po, we have:
  M
 l 
k
M ! 

 pk  1  po  p1  p2  ..  p0 1    
 k 1  m  M  k ! 
k 0
 
1
 M  l k M ! 
 po      
 k 0  m  M  k ! 
 
Elementary Queueing Capacity Analysis 40
Theory
 QUEUEING SYSTEMS
• FINITE CUSTOMER POPULATION, m-SERVER,
FINITE STORAGE: M/M/m/K/M QUEUE
• Consider a situation where we have a finite population of M customers, each
with an “arriving” parameter l.
• In addition, the system has m servers, each with service rate m.
• The system also has finite storage room such that the total numbers of
customers in the system (in queue plus those in service) is no more than K.
• We assume that M≥K≥m; thus customers arriving to find K already in the
system are “lost” and return immediately to the arriving state as if they had just
completed service. This leads to the following set of birth-death coefficients:

l ( M  k ) 0  k  K  1
lk  
 0 else
 km 0  k  m
mk  
mm k m

Elementary Queueing Capacity Analysis 41

Theory
 QUEUEING SYSTEMS
• FINITE CUSTOMER POPULATION, m-SERVER,
FINITE STORAGE: M/M/m/K/M QUEUE
• In the figure below, we see the most complicated of the finite state-
transition-rate diagrams.

Ml (M-1)l (M-m+1)l (M-m+2)l (M-K+2)l

0 1 2 m-2 m-1 m m+1 K-1 K

m 2m m1m mm mm mm

State-transition-rate diagram for m-server, finite-storage, finite-population system, M/M/m/K/M

Elementary Queueing Capacity Analysis 42

Theory
 QUEUEING SYSTEMS
• FINITE CUSTOMER POPULATION, m-SERVER,
FINITE STORAGE: M/M/m/K/M QUEUE
• We consider two separate regions: For 0≤k≤m-1, we have:

l M  i 
k
k 1
li k 1
l M 
pk  po   po   po    , 0  k  m  1
i  0 m i 1 i  0 i  1m m k 
• For the region m≤k≤K, we have:

li
m 1 k 1
l M  i  k 1 l M  i 
p k  po   po  
i  0 m i 1 i 0  
i  1 m i  m mm
k
l  M  k! m  k
 po     m , m  k  K
m  k  m!
Elementary Queueing Capacity Analysis 43
Theory
 QUEUEING SYSTEMS
• FINITE CUSTOMER POPULATION, m-SERVER,
FINITE STORAGE: M/M/m/K/M QUEUE
• In the case of a pure-loss system (that is, M≥K=m), the stationary
state probabilities are given by:

k
 M  l 
  
 k  m 
pk  i
, k  0,1,2,..., m
m
 M  l 
   
i  0  i  m 

• This is known as the Engset distribution.

Elementary Queueing Capacity Analysis 44

Theory