8/10/2019 Queue Networks
1/25
Queuing Networks: BurkesTheorem, KleinrocksApproximation, and Jacksons
Theorem
Wade Trappe
8/10/2019 Queue Networks
2/25
Lectu re Overview
Network of Queues Introduction
Queues in Tandem Product Form Solutions
Burkes Theorem
What is reversibility?
Kleinrocks Approximation Quick Jackson Theorem
8/10/2019 Queue Networks
3/25
In many networking scenarios, a customer or packet must receiveservice from many servers before its final task is completed.
Hence, departures from a queue might become arrivals at anotherqueue.
All that discussion we did in M/G/1 queues becomes veryimportant for networks of queues!
Consider two queues in tandem:
Departures from first queue become arrivals at second queue
First queues arrival process is Poisson with rate andservice time is exponential with rate
Service time at the second queue is exponential with rateand independentof first servers service time
How do we model these two queues together?
Whats the state and state diagram?
Network o f Queues: Setup
2
1
8/10/2019 Queue Networks
4/25
Two Queues in Tandem
Queue2
We need to keep track of N1(t) and N2(t) to describe the
state of the system!
Queue1 1 2
8/10/2019 Queue Networks
5/25
Two Queues in Tandem : State Diagram
Our state diagram must keep track of both N1(t) and N2(t),and the man transitions that are ossible
n1=0 1 2 3
n2=0
1
2
3
2
2
22
2
2 2
2
2 2
2
2
1 1 1
1
111
1 1
8/10/2019 Queue Networks
6/25
Define (N1(t), N2(t)) to be the state vector for the two queues in
tandem
Notice, now, that we have a Markov Process in terms of the statevector
Recall the global balance equations for M/M/1 queues:
What this means is: At steady state, the amount entering a state is
equal to the amount leaving the state.
We similarly may find the global balance equations for this twoqueue system
Global Balance Equations fo r 2 Queues, pg 1
,...2,1j,ppp
pp
1j1jj
10
8/10/2019 Queue Networks
7/25
Global Balance Equations fo r 2 Queues, pg 2
Global Balance Equations: Case 1
Case 2: n>0, m=0
Case 3: m>0, n=0
Case 4: n>0, m>0
1N,0NP0N,0NP 21221
0N,1nNP1N,nNP0N,nNP 21212211
1mN,1NP1mN,0NPmN,0NP 211212212
mN,1nNP
1mN,1nNP
1mN,nNPmN,nNP
21
211
2122121
8/10/2019 Queue Networks
8/25
Steady State PMFs fo r Two M/M/1 Queues
We may show that the following joint probability mass function
satisfies the global balance equations
where ri=/i
So, how do we get P(N1=n)?
Easy, its just an M/M/1!
So P(N1=n) = (1-r1)r1n for n=0,1,2
How do we get P(N2=m)?
Answer: Its a marginal. Integrate out the joint pmf!
Sum over all n to get:
m222 1mNP rr
m22n
1121 11mN,nNP rrrr
Check
This!
8/10/2019 Queue Networks
9/25
Steady State PMFs fo r Two Queues, pg 2
This looks interesting
means
Thus, the number of customers at queue 1 and the number atqueue 2 at a particular time are independentrandom variables!
The steady state at queue 2 is the same as for an M/M/1 queue
with Poisson arrival rate and exponential service time 2.
Definition:A network of queues is said to have a product-form
solutionwhen the joint pmf of the number of customers at eachqueue is the product of the marginal pmfs of the number of
customers at each queue.
m22n1121 11mN,nNP rrrr
mNPnNPmN,nNP 2121
8/10/2019 Queue Networks
10/25
Burkes Theorem
Burkes Theorem is the fundamental result describing product formsolutions
Burkes Theorem:Consider an M/M/1, M/M/m, M/M/infinity queuingsystem at steady state with arrival rate , then
The departure process is Poisson with rate ;
At each time t, the number of customers in the system N(t) is independentof the sequence of departure times prior to t.
What Burkes Theorem implies:
Two queue problem follows from Burkes Thm (arrivals to queue 2 arePoisson with rate ).
Arrivals to queue 2 prior to time t are departures from queue 1 prior totime t, thus Burkes theorem says queue-1s departures (queue-2s
arrivals) are independent of N1(t).
N2(t) is determined by the sequence of arrivals from queue-1 prior to timet and independent of service times, then N1(t) and N2(t) are independentas random variables.
Note: N1(t) and N2(t) are not independent as processes!
8/10/2019 Queue Networks
11/25
Consider the network of queues:
Here Queue 1 is driven by a Poisson process with rate 1,, andthe departures are randomly routed to queues 2 and 3.
Queue 3 has an additional, independent Poisson arrival processwith rate 2.
Example Application of Burkes Thm
1
2
3
1/2
1/2
1
2
8/10/2019 Queue Networks
12/25
Example Application of Burkes Thm, pg 2
Burkes Theorem says: N
1
(t) and N2
(t) are independent
N1(t) and N3(t) are independent
Recall that the random split of a Poisson process yieldsindependent Poisson processes Hence inputs to Queue 2 and Queue 3 are independent
Input to Queue 2 is Poisson with rate 1/2
Input to Queue 3 is Poisson with rate 1/2 + 2
Thus
where r1=1/1, r2=1/22 , r3=(1/2 2/3. All queues areassumed to be stable.
n
33
m
22
k
11321 111n)t(N,m)t(N,k)t(NP rrrrrr
8/10/2019 Queue Networks
13/25
Reversible Markov Processes
In order to prove Burkes theorem, we need the concept of thereversibility of a Markov process.
A stationary Markov process X(t), with a countable state space (i.e. aMarkov chain will do), is reversibleif X(t) and Y(t)=X(-t) have thesame joint distribution at arbitrarily chosen instants {t1, t2, , tN}.
A necessary and sufficient condition for reversibility is
where {pi} and {pij} are the stationary probabilities and transitionprobabilities of X(t)
This condition can be easily shown for M/M/1 queuesbut we willshow it in more general form
In fact, it holds for any birth-death process, and N(-t) is statisticallyidentical to N(t)
i,jjj,ii pppp
8/10/2019 Queue Networks
14/25
Reversible Markov Processes, pg 2
Time-Reversal Theorem:Let {X(t): t>=0} be a stationary
Markov process with (infinitesimal) generator P=[pij], andwith initial distribution equal to stationary distribution. Thenfor all T>0, the time-reversed process
is equivalent to a stationary Markov process with(infinitesimal) generator given by:
for all state pairs (i,j)
Tt0:)tT(X
i
j ij
ij p
ppp~
ijP~
8/10/2019 Queue Networks
15/25
Reversible Markov Processes, pg 3
Proof:Let Q(t)=[qij(t)] denote the transition probabilities of
X(t),
We need to show X(T-t) is a Markov process with transitionprobabilities
Then we obtain the necessary and sufficient condition bydifferentiating this and setting t=0. (Its an infinitesimal
generator)
i)0(X|j)t(XP)t(qij
i
j ij
ijp
)t(qp)t(q~)t(Q
~
8/10/2019 Queue Networks
16/25
Reversible Markov Processes, pg 4
Consider the interval (0,t+s] and divide it into (0,t] and
(t,t+s], i.e. set T=t+s.
The joint probability of the three random variables
is
Similarly we have
)0(X)st(X~
)s(X)t(X~
)st(X)0(X~
)t(q)s(qp
i)st(X,i)s(X,i)0(XP
i)st(X~,i)t(X~,i)0(X~P
01122 i,ii,ii
012
210
)t(qp
i)ts(X,i)s(XPi)t(X~,i)0(X
~P
011 i,ii
0110
8/10/2019 Queue Networks
17/25
Reversible Markov Processes, pg 5
The conditional probability
Hence, we have shown that is a Markov process with
generator (after differentiation).P~
j,i
12
i
i,ii
012
q~i)t(X~|i)st(X~P
p
)t(qp
i)0(X~,i)t(X
~|i)st(X
~P
1
122
tX~
8/10/2019 Queue Networks
18/25
B-D Processes are Reversib le
It is now easy to show the following
Time Reversibility of B-D Processes:The stationary B-D process N(t)with generator Pand steady state probabilities pis a time-reversibleMarkov process. Thus, the time-reversal of the death process is a birth
process.
Burkes Theorem follows from this:
Interdeparture times of the forward-time system are the interarrivaltimes of the time-reversed system... Hence we have Poisson withrate coming out of the system.
Fix a time t, then the departures before time t from the forwardsystem are arrivals after time t in the reverse system.
Arrivals in reverse system are Poisson, and thus arrivals in reversesystem after time t do not depend on N(t)
Consequently, departures after time t in forward system do notdepend on N(t)
8/10/2019 Queue Networks
19/25
8/10/2019 Queue Networks
20/25
Network o f Queues, M/D/1 f irst queue
The interarrival times of the second queue must be at least 1/
Why?
Now, each packet arriving at either queue takes 1/time toprocess.
The first packet being finished by first queue is immediately
sent to second server
It takes at leastanother 1/amount of time for first queue to
get and finish the next packet/customer.
So, first packet will be finished by second server at or before
the next packet arrives to second server.
Result: No queue (waiting) at second system!
8/10/2019 Queue Networks
21/25
Two Queues, co rrelated service t imes
Earlier we considered the service times independent of each other
and independent of the arrival times. Reality:A big packet at the first system is probably still a big
packet at the second system!
Interarrival times at the second queue are strongly correlatedwith packet lengths!
Long packets at first system will typically find the queue at thesecond server more empty
Shorter packets from the first system will typically find the queueat the second server more busy because the second server is
processing some prior big packet
It is tough to find an analytical solution for joint pmf underdependence assumptions!
8/10/2019 Queue Networks
22/25
The Kleinroc k Independence Approximat ion
We have argued that in practice there is dependence upon the
interarrival times and service times. Independence is lost after the first system!
Reality hurt us, but reality provides us one more gift
Reality:Real networks typically involve more than one stream of
packets merging at a node The combination of multiplestreams helps restoreindependence in many cases!
This observation is due to Kleinrock.
Kleinrocks Approximation:It is often appropriate to useM/M/1 queues for each communication link when the arrivals atentry points are Poisson, packet lengths are roughly exponentially
distributed, network is dense and traffic is heavy.
8/10/2019 Queue Networks
23/25
Quick Look at Jacksons Theorem
Many queuing networks, a packet/customer may visit a queue
more than once. Burkes theorem does not apply!
Typical example: Queue with feedback
If the arrival rate is much less than departure rate, then netarrival process has a few, isolated external arrivals followed bya burst of feedback arrivals (dependent on packet length).
a
p
1-p
8/10/2019 Queue Networks
24/25
Jacksons Theorem, (Open) pg 1.
Consider a queuing network consisting of M separate service stations,each with its own queue.
Define the vector process:
In an open queuing network, customers may arrive from an externalsource and eventually may leave the network (as depicted in previousslide).
A closed network has no arrivals or departures from the system (totalcustomers is fixed just they may move around)
Assumptions: The rate of the source (birth process) is , and a customer goes to
station i with probability qsi.
Service time at station k is exponential with rate k.
Customers are routed according to a Markov chain: Probabilitythat a customer departing station i goes to station j is qij.
)t(N,),t(N),t(NtN M21
8/10/2019 Queue Networks
25/25
Jacksons Theorem, (Open) pg 2.
Jacksons Decomposition Theorem:For an open queue as
described, the joint distribution of the queue vector n(t) isgiven by:
where P(Nk=nk) is the steady state pmf of an M/M/1 systemwith arrival rate kand service rate k, i.e.
Where rkdescribes the utilization factor of system k.
KK2211 nNPnNPnNPnP
kn
kkk 1nNP k rr