Scribd
Upload
4 views12 pages

OR I Lecture - Matl 03 02 2021

Uploaded by

SHUBHAM SAHA
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
4 views12 pages

OR I Lecture - Matl 03 02 2021

Uploaded by

SHUBHAM SAHA
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 12

QUEUING THEORY

(Part--I)…contd.
(Part

Dr. Prasun Das


Indian Statistical Institute
SQC & OR Unit
Kolkata, India
[email protected]

3-Feb-21 Indian Statistical Institute 1


Poisson Process: Definition
• Let N(t) be the no. of arrivals in time interval (0,t) with rate  i.e., time
between successive intervals is exponentially distributed with
parameter  and independent of past.

• Then, N(t) has a Poisson process with parameter .

• P(N(t) = n) = e-t(t)n , E[N(t)] = t


n!

From this, we can find out the probability of P(N(∆t)) where ∆t is small
(   t ) 2
Pr( 0 )  e    t  1    t  ...  1    t  o (  t )  Pr( 0 )  1    t
2!
(   t ) 2
Pr( 1)    te   t    t [1    t  ...]    t  o (  t )  Pr( 1)    t
2!
Pr(  2 )  ...  0
3-Feb-21 Indian Statistical Institute 2
Poisson Process
• {N(t), t0} : Stochastic (time indexed r.v.) arrival process with N(0) = 0. Then,

 P(1) = Prob {an arrival occurs between t and t+t} = .t + o(t)

 P(>1) = Prob {more than one arrival occurs between t and t+t} = o(t)

 No. of arrivals are independent in non-overlapping intervals, dept. only on t

• In other words, P(N(t) = n) = pn(t) = prob. of n arrivals in time t

 Poisson Distrn. with mean of t arrivals (or, mean arrival rate of ).

3-Feb-21 Indian Statistical Institute 3


Poisson Process
• Let T be the r.v. “time between successive arrivals”

• Then, Pr{T t}: Pr. {no arrival in time t} = p0(t) = e - t

 CDF of T: F(T) = Pr {T t} = 1 - e - t

 PDF of T: f(t) = F/(t) = .e - t

 T follows Exponential Distrn. with mean time between arrivals as 1/.

Inter-arrival time: Exponential  No. of arrivals: Poisson

3-Feb-21 Indian Statistical Institute 4


(M)arkovian Property: Exponential Distribution
• Suppose, Service time ~ exponential distrn.

• Property: The probability that a customer currently in service has t units of


remaining service is independent of how long (aging) it has already been in
service.

• In other words, Pr{T t1 | T t0} = Pr{0  T  t1 - t0} = 1 - e - (t1 - t0)

• The Exponential Distrn. is the only continuous distrn. which exhibits this
memoryless property.

• This property aids in a relatively simple analysis of queuing problems under


Poisson (arrival)-exponential (service) characteristics.

3-Feb-21 Indian Statistical Institute 5


Stochastic Process
• A Stochastic Process is a family of random variables {X(t), tT} defined over
some index set/parameter space T (time range). X(t) denotes the state of the
process at time t. The set of possible states constitutes the state space.

• If T is countable (for ex, T = {0,1,2,…}), {X(t), tT} is said to be a discrete-


parameter stochastic process defined on the index set T.

• If T is an interval (for ex, T = {t: 0<t<}), {X(t), tT} is said to be a


continuous-parameter stochastic process defined on the index set T.

• A discrete or continuous parameter stochastic process is said to be a Markov


process if, for any set of n time points t1 < t2 …< tn in T, the conditional distrn.
of X(tn), given the values of X(t1), X(t2), X(t3),…,X(tn-1), depend only on the
immediately preceding value, X(tn-1). Here, the state space is always
continuous.
3-Feb-21 Indian Statistical Institute 6
Markov Process: Memoryless
• In otherwords, Pr{X(tn) xn | X(t1)=x1,…, X(tn-1)=xn-1} = Pr{X(tn) xn | X(tn-1)=xn-1}

• The ‘future’ condition of the process, given the ‘present’ condition, is


independent of the ‘past’, and the process is thus ‘memoryless’.

• If the state space is discrete, the Markov process is generally called a


Markov chain.

• A discrete-parameter Markov process with discrete state space is simply a


Markov chain.

• A continuous-parameter Markov process with discrete state space is


simply a continuous-parameter Markov chain.

3-Feb-21 Indian Statistical Institute 7


Markov Process & Markov Chain

Parameter Space (T)


State Space: X(t) Discrete Continuous
Discrete Markov Chain Cont. Parameter
Markov Chain
Continuous Markov Process Markov Process

Question: Queuing Model: ??

3-Feb-21 Indian Statistical Institute 8


Queuing Models: Markov chain
• Poisson process is a stochastic process.

• State space of any queue: discrete (no. of customers present which is non-
negative integers)

• So, a large percentage of queuing problems are categorized as continuous-


parameter Markov chain.

• Many such models have additional birth (arrival)-death (service) property


that the net change across an infinitesimal time interval can never be other
than -1, 0, or +1.
 
n-1 n n+1
 

3-Feb-21 Indian Statistical Institute 9


Poisson Queue
• Discrete State space: n (no. of customers at time t)

• Arrival (birth) Rate: n

• Service (death) Rate: n

• pn(t): Prob. that n customers in the system at time t. Then,

pn(t+t) = pn(t).[1 - n.t ].[1 - n.t ] + pn-1(t).[n-1.t].[1 - n-1.t ]


+ pn+1(t).[1 - n+1.t ].[ n+1.t ] + o(t)2

• Differential-Difference Equation:

dpn t  pn t  t   pn t 
  n   n  pn t   pn 1 t .n 1  pn 1 t . n 1
dt t
dp0 t 
 0 . p0 t   1. p1 t 
dt

3-Feb-21 Indian Statistical Institute 10


Poisson Queue
• These equations may be solved with the proper initial conditions to get the
Transient Solutions.

• If the queue starts with n in the system, then the corresponding initial
conditions would be:

pi(0) = 0, for in and pn(0) = 1, for i=n

• Steady state (equilibrium) condition:


dpn t 
 0 and lt pn (t )  pn
dt t 

 0  n   n  pn  n 1. pn 1   n 1. pn 1  pn 1 


n   n n 1
pn  pn 1 , n  1
 n 1  n 1
0  0 . p0  1. p1 0
p1  . p0 , n  0
1

3-Feb-21 Indian Statistical Institute 11


Poisson Queue
n   n n 1
pn 1  pn  pn 1 , n  1
 n 1  n 1
0
p1  . p0 , n  0
1
0 .1 0 .1...n 1
p2  . p0 …………....... pn  . p0
1. 2 1. 2 ... n

1

  n
i 1 
Since  pn  1, p0  1     S 1 , say Check!!
n 0  n1 i 1 i 

So, a n+s condition for the existence of a steady-state solution is the


convergence of the infinite series S.

3-Feb-21 Indian Statistical Institute 12

You might also like