Stat 150 Stochastic Processes

Spring 2009

Lecture 24: Queuing Models

Lecturer: Jim Pitman

Queueing Models The simplest setup is a single server with some service rules takes a random to process then the customer departs.
h h h h h

h h h h

Queue h h h

Generically, X (t) = # of customers in system(queue) at time t, and model for evolution of X . M/M/1 M = Markov First M is for input stream Second M is for service process 1 = # of servers Poisson arrival at rate Independent exponential service time with rate Each customer that is served takes exp() time The service times T1 , T2 , . . . are i.i.d. exp() random variables independent of the arrival stream

Lecture 24: Queuing Models

2 # " ! 1 # " ! 0 " !

. . . . . . 3 #

qij = rate of transition from i to j = o diagonal elements of the A matrix q0 = = A00 q01 = q0j = 0, for all j = 0, j = 1 q1 = + = A11 q10 = q12 = q1j = 0, for all j = 0, j = 1, j = 2 Condition for stability of the queue : < = is exceptional (null-recurrent chain). > is linear growth (like a RW with upward drift: transient chain). Stability means lim P(Xt = j ) = j > 0 with
t j

j = 1.

We can calculate . How? Pt = , for all t

d and evaluate at 0 = A = 0. (Note: we did not try to solve dierential Take dt equations for Pt . In the M/M/1 queue Pt (i, j ) is a nasty power series (involving Bessel functions))

Use We get

qi = Aii ,

qij = Aij , i = j,

and
i

i Aij = 0, for all j

j,
i=j

i qij j qj = 0 i qij = j qj
i=j

=j,

Lecture 24: Queuing Models

That is , at equilibrium, Rate in to j = Rate out of j Fact: For birth/death chain on integers, the only possible equilibrium is reversible. So always look for solution of Rev EQ. 0 = 1 n 1 = 2 = = n = 0 ( )n = n1 . . .

Use
n=0

n = 1, get n = ( )n (1 ),

n = 0, 1, 2, . . .

Check P549(2.9) In the limit, mean queue length = 1/(1 ) = . If is close to , this is . large. Fix , as , M/M/ Input rate Unlimited number of servers, each operate at rate on one customer at a time
. . . . . . 3 # 3 2 " # ! 2 1 # " ! 0 " !

Use the fact that mean of exp(1 ), exp(2 ), . . . , exp(n ) is exp(1 +2 + +n ). Condition for stability? No matter what > 0, > 0, the system is stable.To see this, establish a Rev EQ probablilty distribution:

Lecture 24: Queuing Models

0 = 1 1 = 2 2 2 = 3 3 . . .

n 1 = = n = 0 ( )n = n1 n n!

So 0 = e/ . The equilibrium probability distribution is Poisson(/) and the mean queue length is /.