Analyzing the M/G/1 Queue with a Branching Process
Céline Comte Nokia Bell Labs France - Télécom ParisTech
Reading Group ”Network Theory” December 18, 2017 M/G/1 queue
Branching process
Stability Recurrence and transcience Positive recurrence
Related work and references
2 © 2017 Nokia Public M/G/1 queue
Branching process
Stability Recurrence and transcience Positive recurrence
Related work and references
3 © 2017 Nokia Public The M/G/1 queue with FCFS service discipline
λ µ
M The arrival process is Poisson → Arrival rate 0 < λ < +∞ G The service times are i.i.d. with a general distribution 1 → Mean service time 0 < µ < +∞ 1 A single server ? Infinite queue length
FCFS Customers leave in their arrival order
4 © 2017 Nokia Public • The number of arrivals during a time interval of length t has a Poisson distribution with mean λt. • The numbers of arrivals during two disjoint intervals are independent.
More on Poisson processes
t1 Time
• Exponentially distributed inter-arrival times
5 © 2017 Nokia Public • The numbers of arrivals during two disjoint intervals are independent.
More on Poisson processes
t1 Time
• Exponentially distributed inter-arrival times • The number of arrivals during a time interval of length t has a Poisson distribution with mean λt.
5 © 2017 Nokia Public • The numbers of arrivals during two disjoint intervals are independent.
More on Poisson processes
tt1 Time
• Exponentially distributed inter-arrival times • The number of arrivals during a time interval of length t has a Poisson distribution with mean λt.
5 © 2017 Nokia Public More on Poisson processes
tt1 Time
• Exponentially distributed inter-arrival times • The number of arrivals during a time interval of length t has a Poisson distribution with mean λt. • The numbers of arrivals during two disjoint intervals are independent.
5 © 2017 Nokia Public More on Poisson processes
′ tt1 t Time
• Exponentially distributed inter-arrival times • The number of arrivals during a time interval of length t has a Poisson distribution with mean λt. • The numbers of arrivals during two disjoint intervals are independent.
5 © 2017 Nokia Public The M/G/1 queue with FCFS service discipline
λ µ
M The arrival process is Poisson → Arrival rate 0 < λ < +∞ G The service times are i.i.d. with a general distribution 1 → Mean service time 0 < µ < +∞ 1 A single server ? Infinite queue length
FCFS Customers leave in their arrival order
6 © 2017 Nokia Public Queue state
• Xt = number of customers in the queue at time t
• (Xt)t≥0 is not a Markov process (in general)
3
2
Queue length 1
0 0 2 4 6 8 10 12 14 Time
7 © 2017 Nokia Public Departure time Dn
• Dn = departure time of the n-th customer
• Dn is a regeneration point of (Xt)t≥0
3
2
Queue length 1
0 0 2 4 6 8 10 12 14 Time
8 © 2017 Nokia Public Departure time Dn
• Dn = departure time of the n-th customer
• Dn is a regeneration point of (Xt)t≥0
3
2
Queue length 1
0 D0 D1 D2 D3 D4 D5 Time
8 © 2017 Nokia Public Chain embedded at departure times
• = Yn XDn number of customers left behind customer n • (Yn)n∈N is a Markov chain
3
2
Queue length 1
0 D0 D1 D2 D3 D4 D5 Time
9 © 2017 Nokia Public Chain embedded at departure times
• Because of the FCFS assumption, Yn = number of customers that arrived since customer n entered the queue
3
2
Queue length 1
0 D0 D1 D2 D3 D4 D5 Time
10 © 2017 Nokia Public Chain embedded at departure times
• Because of the FCFS assumption, Yn − Yn−1 + 1 = number of customers that arrived between the departures of customers n − 1 and n
3
2
Queue length 1
0 D0 D1 D2 D3 D4 D5 Time
11 © 2017 Nokia Public Chain embedded at departure times
• Because of the FCFS assumption, Yn − Yn−1 + 1 = number of customers that arrived during the service of customer n
3
2
Queue length 1
0 D0 D1 D2 D3 D4 D5 Time
11 © 2017 Nokia Public Chain embedded at departure times
• Because of the FCFS assumption, Yn − (Yn−1 − 1)+ = number of customers that arrived during the service of customer n
3
2
Queue length 1
0 D0 D1 D2 D3 D4 D5 Time
11 © 2017 Nokia Public p2 p1 p2
p0 p1 p0
• The number of arrivals during a service time of length t has a Poisson distribution with mean λt. • Letting F denote the CDF of the service time distribution, ∫ +∞ (λt)k = −λt pk e dF(t). 0 k! • Independence
Why is it a Markov chain ?
0 1 2 3 ...4
• = pk probabily that k customers arrive during the service = pk time of a customer
12 © 2017 Nokia Public p2 p1 p2
p0 p1 p0
• Letting F denote the CDF of the service time distribution, ∫ +∞ (λt)k = −λt pk e dF(t). 0 k! • Independence
Why is it a Markov chain ?
0 1 2 3 ...4
• = pk probabily that k customers arrive during the service = pk time of a customer • The number of arrivals during a service time of length t has a Poisson distribution with mean λt.
12 © 2017 Nokia Public p2 p1 p2
p0 p1 p0
• Independence
Why is it a Markov chain ?
0 1 2 3 ...4
• = pk probabily that k customers arrive during the service = pk time of a customer • The number of arrivals during a service time of length t has a Poisson distribution with mean λt. • Letting F denote the CDF of the service time distribution, ∫ +∞ (λt)k = −λt pk e dF(t). 0 k!
12 © 2017 Nokia Public p2 p1 p2
p0 p1 p0
Why is it a Markov chain ?
0 1 2 3 ...4
• = pk probabily that k customers arrive during the service = pk time of a customer • The number of arrivals during a service time of length t has a Poisson distribution with mean λt. • Letting F denote the CDF of the service time distribution, ∫ +∞ (λt)k = −λt pk e dF(t). 0 k! • Independence
12 © 2017 Nokia Public p2 p1
p0
Why is it a Markov chain ?
p 0 1 2 2 3 ...4 p1 p0
• = pk probabily that k customers arrive during the service = pk time of a customer • The number of arrivals during a service time of length t has a Poisson distribution with mean λt. • Letting F denote the CDF of the service time distribution, ∫ +∞ (λt)k = −λt pk e dF(t). 0 k! • Independence
12 © 2017 Nokia Public Why is it a Markov chain ?
p p2 p 0 1 1 2 2 3 ...4 p0 p1 p0
• = pk probabily that k customers arrive during the service = pk time of a customer • The number of arrivals during a service time of length t has a Poisson distribution with mean λt. • Letting F denote the CDF of the service time distribution, ∫ +∞ (λt)k = −λt pk e dF(t). 0 k! • Independence
12 © 2017 Nokia Public Why is it a Markov chain ?
p p2 p 0 1 1 2 2 3 ...4 p0 p1 p0
• = pk probabily that k customers arrive during the service = pk time of a customer • Transition matrix : p p p p ... p ... 0 1 2 3 k p p p p ... p ... 0 1 2 3 k p0 p1 p2 ... pk ... p0 p1 ... pk ... p0 ... pk ...
13 © 2017 Nokia Public • Aperiodic
→ If (Yn)n∈N is positive recurrent, then it is also ergodic
Properties of (Yn)n∈N
p p2 p 0 1 1 2 2 3 ...4 p0 p1 p0
• Irreducible
→ All the states of (Yn)n∈N have the same nature (positive recurrent, null recurrent or transcient) → We just need to know the nature of state 0
14 © 2017 Nokia Public Properties of (Yn)n∈N
p p2 p 0 1 1 2 2 3 ...4 p0 p1 p0
• Irreducible
→ All the states of (Yn)n∈N have the same nature (positive recurrent, null recurrent or transcient) → We just need to know the nature of state 0 • Aperiodic
→ If (Yn)n∈N is positive recurrent, then it is also ergodic
14 © 2017 Nokia Public Stability of (Yn)n∈N
p p2 p 0 1 1 2 2 3 ...4 p0 p1 p0
• State 0 is recurrent ⇔ If the queue starts empty, then with probability 1 it empties out an infinite number of times ⇔ If the queue starts empty, then with probability 1 it eventually empties out
15 © 2017 Nokia Public Stability of (Yn)n∈N
p p2 p 0 1 1 2 2 3 ...4 p0 p1 p0
Assume that state 0 is recurrent. • State 0 is positive recurrent ⇔ If the queue starts empty, then the mean time until it empties out again is finite • State 0 is null recurrent ⇔ If the queue starts empty, then the mean time until it empties out again is infinite
16 © 2017 Nokia Public Busy period
• We just need to look at one busy period of the queue.
3
2
Queue length 1
0 Start of a End of a busy period busy period
17 © 2017 Nokia Public M/G/1 queue
Branching process
Stability Recurrence and transcience Positive recurrence
Related work and references
18 © 2017 Nokia Public − random, − with a probability distribution that is the same for all nodes, − independent of the number of children of other nodes,
• One node at generation 0 • The children of the nodes of generation n belong to generation n + 1 • The number of children of a node is
• Mean number of children per node ρ < +∞
Definition
• Random tree
A = 2
19 © 2017 Nokia Public − random, − with a probability distribution that is the same for all nodes, − independent of the number of children of other nodes,
• The children of the nodes of generation n belong to generation n + 1 • The number of children of a node is
• Mean number of children per node ρ < +∞
Definition
• Random tree • One node at generation 0
A = 2
19 © 2017 Nokia Public − random, − with a probability distribution that is the same for all nodes, − independent of the number of children of other nodes,
• The number of children of a node is
• Mean number of children per node ρ < +∞
Definition
• Random tree • One node at generation 0 • The children of the nodes of generation n belong to generation n + 1
A = 2
19 © 2017 Nokia Public − random, − with a probability distribution that is the same for all nodes, − independent of the number of children of other nodes, • Mean number of children per node ρ < +∞
Definition
• Random tree • One node at generation 0 • The children of the nodes of generation n belong to generation n + 1 • The number of children of a node is
A = 2
19 © 2017 Nokia Public − with a probability distribution that is the same for all nodes, − independent of the number of children of other nodes, • Mean number of children per node ρ < +∞
Definition
• Random tree • One node at generation 0 • The children of the nodes of generation n belong to generation n + 1 • The number of children of a node is − random,
A = 2
19 © 2017 Nokia Public − independent of the number of children of other nodes, • Mean number of children per node ρ < +∞
Definition
• Random tree • One node at generation 0 • The children of the nodes of generation n belong to generation n + 1 • The number of children of a node is − random, − with a probability distribution that is the same for all nodes, A = 2
19 © 2017 Nokia Public • Mean number of children per node ρ < +∞
Definition
• Random tree • One node at generation 0 • The children of the nodes of generation n belong to generation n + 1 • The number of children of a node is − random, − with a probability distribution that is the same for all nodes, − independent of the number of A = 2 children of other nodes,
19 © 2017 Nokia Public Definition
• Random tree • One node at generation 0 • The children of the nodes of generation n belong to generation n + 1 • The number of children of a node is − random, − with a probability distribution that is the same for all nodes, − independent of the number of A = 2 children of other nodes, • Mean number of children per node ρ < +∞
19 © 2017 Nokia Public Definition
B Given the number of children of the root, the subtrees rooted at these nodes are independent and have the same distribution as the initial tree
A = 2
20 © 2017 Nokia Public Definition
B Given the number of children of the root, the subtrees rooted at these nodes are independent and have the same distribution as the initial tree
A = 2
20 © 2017 Nokia Public Build the tree
21 © 2017 Nokia Public Build the tree
21 © 2017 Nokia Public Build the tree
21 © 2017 Nokia Public Build the tree
21 © 2017 Nokia Public Build the tree
21 © 2017 Nokia Public Build the tree
21 © 2017 Nokia Public Build the tree
21 © 2017 Nokia Public Build the tree
21 © 2017 Nokia Public Build the tree
21 © 2017 Nokia Public Build the tree
21 © 2017 Nokia Public Build the tree
21 © 2017 Nokia Public Build the tree
21 © 2017 Nokia Public Build the tree
21 © 2017 Nokia Public Build the tree
21 © 2017 Nokia Public Build the tree
21 © 2017 Nokia Public Build the tree
21 © 2017 Nokia Public Build the tree
21 © 2017 Nokia Public Build the tree
21 © 2017 Nokia Public Queue fluctuation with time
4
3
2 Queue length
1
0 0 10 20 30 40 50 60 Time
22 © 2017 Nokia Public Building the tree Queue length
Time
23 © 2017 Nokia Public Building the tree Queue length
Time
23 © 2017 Nokia Public Building the tree Queue length
Time
23 © 2017 Nokia Public Building the tree Queue length
Time
23 © 2017 Nokia Public Building the tree Queue length
Time
23 © 2017 Nokia Public Building the tree Queue length
Time
23 © 2017 Nokia Public Building the tree Queue length
Time
23 © 2017 Nokia Public Building the tree Queue length
Time
23 © 2017 Nokia Public Building the tree Queue length
Time
23 © 2017 Nokia Public Building the tree Queue length
Time
23 © 2017 Nokia Public Building the tree Queue length
Time
23 © 2017 Nokia Public Building the tree Queue length
Time
23 © 2017 Nokia Public Realization
4
3
2 Queue length
1
0 0 10 20 30 40 50 60 Time
24 © 2017 Nokia Public Building the tree Queue length
Time
25 © 2017 Nokia Public Building the tree Queue length
Time
25 © 2017 Nokia Public Building the tree Queue length
Time
25 © 2017 Nokia Public Building the tree Queue length
Time
25 © 2017 Nokia Public Building the tree Queue length
Time
25 © 2017 Nokia Public Building the tree Queue length
Time
25 © 2017 Nokia Public Building the tree Queue length
Time
25 © 2017 Nokia Public Building the tree Queue length
Time
25 © 2017 Nokia Public Building the tree Queue length
Time
25 © 2017 Nokia Public Building the tree Queue length
Time
25 © 2017 Nokia Public Building the tree Queue length
Time
25 © 2017 Nokia Public Building the tree Queue length
Time
25 © 2017 Nokia Public Building the tree Queue length
Time
25 © 2017 Nokia Public Building the tree Queue length
Time
25 © 2017 Nokia Public Building the tree Queue length
Time
25 © 2017 Nokia Public Building the tree Queue length
Time
25 © 2017 Nokia Public Building the tree Queue length
Time
25 © 2017 Nokia Public Building the tree Queue length
Time
25 © 2017 Nokia Public Why is it a branching process ?
26 © 2017 Nokia Public Definition
• Random tree • One node at generation 0 • The children of the nodes of generation n belong to generation n + 1 • The number of children of a node is − random, − with a probability distribution that is the same for all nodes, − independent of the number of A = 2 children of other nodes, • Mean number of children per node ρ < +∞
27 © 2017 Nokia Public • = pk probability that k customers arrive = pk during a single service time = pk probability that a node has k children • The number of arrivals during a service time of length t has a Poisson distribution with mean λt. • Independence
t A = 2 Time
Why is it a branching process ?
28 © 2017 Nokia Public • The number of arrivals during a service time of length t has a Poisson distribution with mean λt. • Independence
t Time
Why is it a branching process ?
• = pk probability that k customers arrive = pk during a single service time = pk probability that a node has k children
A = 2
28 © 2017 Nokia Public • Independence
Why is it a branching process ?
• = pk probability that k customers arrive = pk during a single service time = pk probability that a node has k children • The number of arrivals during a service time of length t has a Poisson distribution with mean λt.
t A = 2 Time
28 © 2017 Nokia Public Why is it a branching process ?
• = pk probability that k customers arrive = pk during a single service time = pk probability that a node has k children • The number of arrivals during a service time of length t has a Poisson distribution with mean λt. • Independence
t A = 2 Time
28 © 2017 Nokia Public Mean number of children per node
• S = service time of a given customer. 1 S is a random variable with mean µ . • A = number of customers arrived during A = the service of this customer • Given S, A has a Poisson distribution with mean λS.
S A = 2 Time
29 © 2017 Nokia Public Mean number of children per node
• Mean number of children of a node E(A) = E(E(A|S)) = E(λS) = λE(S) λ = µ . λ ⇒ ρ = µ = load of the queue, λ ρ = µ = mean number of children of λ A = 2 ρ = µ = a node in the tree.
30 © 2017 Nokia Public M/G/1 queue
Branching process
Stability Recurrence and transcience Positive recurrence
Related work and references
31 © 2017 Nokia Public M/G/1 queue
Branching process
Stability Recurrence and transcience Positive recurrence
Related work and references
32 © 2017 Nokia Public Recurrence and transcience
λ • ρ = µ = load of the queue, λ ρ = µ = mean number of children per node.
• Branching process result : − ρ ≤ 1 : the tree dies out with probability 1. ρ ≤ 1 : (assuming that p1 < 1 if ρ = 1) − ρ > 1 : the tree dies out with probability < 1. • Queueing reformulation :
− ρ ≤ 1 : (Yn)n∈N is recurrent. − ρ > 1 : (Yn)n∈N is transcient.
33 © 2017 Nokia Public • P = probability that the queue empties P = probability that the tree is finite
• P satisfies the fixed-point equation
+∞∑ = k P P pk. k=0
Sketch of proof
• = pk probability that k customers arrive = pk during a single service time = pk probability that a node has k children
N2 = 2
34 © 2017 Nokia Public • P satisfies the fixed-point equation
+∞∑ = k P P pk. k=0
Sketch of proof
• = pk probability that k customers arrive = pk during a single service time = pk probability that a node has k children
• P = probability that the queue empties P = probability that the tree is finite
N2 = 2
34 © 2017 Nokia Public Sketch of proof
• = pk probability that k customers arrive = pk during a single service time = pk probability that a node has k children
• P = probability that the queue empties P = probability that the tree is finite
• P satisfies the fixed-point equation N2 = 2 +∞∑ = k P P pk. k=0
34 © 2017 Nokia Public Sketch of proof
• = pk probability that k customers arrive = pk during a single service time = pk probability that a node has k children
• P = probability that the queue empties P = probability that the tree is finite
• P satisfies the fixed-point equation N2 = 2 +∞∑ = k P P pk. k=0
34 © 2017 Nokia Public • P satisfies the fixed-point equation
P = ϕ(P),
where +∞∑ ϕ = k (x) x pk k=0
is the generating function of (pk)k∈N
Sketch of proof
• P satisfies the fixed-point equation
+∞∑ = k P P pk k=0
N2 = 2
35 © 2017 Nokia Public Sketch of proof
• P satisfies the fixed-point equation
+∞∑ = k P P pk k=0
• P satisfies the fixed-point equation
P = ϕ(P),
where +∞∑ N2 = 2 ϕ = k (x) x pk k=0
is the generating function of (pk)k∈N
35 © 2017 Nokia Public Sketch of proof ρ = 1.4 1 y = ϕ(x) y = x 0.8 Slope in 1 = ρ
0.6 y
0.4 +∞∑ ′ϕ(x) = xkp p0 k k=0 +∞∑ 0.2 ϕ′ = k−1 (x) kx pk k=1 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 x
36 © 2017 Nokia Public Sketch of proof
• P is the smallest solution in [0,1] of the fixed-point equation
P = ϕ(P),
where +∞∑ ϕ = k (x) x pk k=0
is the generating function of (pk)k∈N N2 = 2
37 © 2017 Nokia Public • P = lim Pn n→+∞
• (Pn)n∈N satisfies +∞∑ = k = ϕ Pn+1 Pn pk, that is, Pn+1 (Pn) k=0
• P0 = 0
Sketch of proof
• Pn = probability that the population Pn = is extincted at generation n
N2 = 2
38 © 2017 Nokia Public • (Pn)n∈N satisfies +∞∑ = k = ϕ Pn+1 Pn pk, that is, Pn+1 (Pn) k=0
• P0 = 0
Sketch of proof
• Pn = probability that the population Pn = is extincted at generation n
• P = lim Pn n→+∞
N2 = 2
38 © 2017 Nokia Public ( ) +∞∪ { } = P the population is extincted at generation n n=0 ( ) = lim P population is extincted n→+∞ at generation n
• (Pn)n∈N satisfies +∞∑ = k = ϕ Pn+1 Pn pk, that is, Pn+1 (Pn) k=0
• P0 = 0
Sketch of proof
• Pn = probability that the population Pn = is extincted at generation n
• P = lim Pn n→+∞ P(the population is finite)
N2 = 2
38 © 2017 Nokia Public ( ) = lim P population is extincted n→+∞ at generation n
• (Pn)n∈N satisfies +∞∑ = k = ϕ Pn+1 Pn pk, that is, Pn+1 (Pn) k=0
• P0 = 0
Sketch of proof
• Pn = probability that the population Pn = is extincted at generation n
• P = lim Pn n→+∞ P(the( population is finite) ) +∞∪ { } = P the population is extincted at generation n n=0
N2 = 2
38 © 2017 Nokia Public • (Pn)n∈N satisfies +∞∑ = k = ϕ Pn+1 Pn pk, that is, Pn+1 (Pn) k=0
• P0 = 0
Sketch of proof
• Pn = probability that the population Pn = is extincted at generation n
• P = lim Pn n→+∞ P(the( population is finite) ) +∞∪ { } = P the population is extincted at generation n n=0 ( ) = P population is extincted →+∞lim at generation n n N2 = 2
38 © 2017 Nokia Public • (Pn)n∈N satisfies +∞∑ = k = ϕ Pn+1 Pn pk, that is, Pn+1 (Pn) k=0
• P0 = 0
Sketch of proof
• Pn = probability that the population Pn = is extincted at generation n
• P = lim Pn n→+∞
N2 = 2
38 © 2017 Nokia Public • P0 = 0
Sketch of proof
• Pn = probability that the population Pn = is extincted at generation n
• P = lim Pn n→+∞
• (Pn)n∈N satisfies +∞∑ k + = , + = ϕ( ) Pn 1 Pn pk that is, Pn 1 Pn = k=0 N2 2
38 © 2017 Nokia Public Sketch of proof
• Pn = probability that the population Pn = is extincted at generation n
• P = lim Pn n→+∞
• (Pn)n∈N satisfies +∞∑ k + = , + = ϕ( ) Pn 1 Pn pk that is, Pn 1 Pn = k=0 N2 2
• P0 = 0
38 © 2017 Nokia Public Sketch of proof
• P is a solution of the fixed-point equation
P = ϕ(P)
• It is also the limit of the sequence (Pn)n∈N defined recursively by P0 = 0 and
Pn+1 = ϕ(Pn), ∀n ∈ N.
N2 = 2
39 © 2017 Nokia Public Sketch of proof ρ = 1.4 1 y = ϕ(x) y = x 0.8 Slope in 1 = ρ
0.6 y
0.4 +∞∑ ′ϕ(x) = xkp p0 k k=0 +∞∑ 0.2 ϕ′ = k−1 (x) kx pk k=1 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 x
40 © 2017 Nokia Public Sketch of proof ρ = 0.6 1 Slope y = ϕ(x) y = x in 1 = ρ 0.8
0.6 p0 y
0.4 +∞∑ ′ϕ = k (x) x pk k=0 +∞∑ 0.2 ϕ′ = k−1 (x) kx pk k=1 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 x
41 © 2017 Nokia Public Sketch of proof ρ = 1 1 y = ϕ(x) y = x Slope 0.8 in 1 = ρ
0.6 y p0 0.4 +∞∑ ′ϕ = k (x) x pk k=0 +∞∑ 0.2 ϕ′ = k−1 (x) kx pk k=1 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 x
42 © 2017 Nokia Public Recurrence and transcience
λ • ρ = µ = load of the queue, λ ρ = µ = mean number of children per node.
• Branching process result : − ρ ≤ 1 : the tree dies out with probability 1. ρ ≤ 1 : (assuming that p1 < 1 if ρ = 1) − ρ > 1 : the tree dies out with probability < 1. • Queueing reformulation :
− ρ ≤ 1 : (Yn)n∈N is recurrent. − ρ > 1 : (Yn)n∈N is transcient.
43 © 2017 Nokia Public M/G/1 queue
Branching process
Stability Recurrence and transcience Positive recurrence
Related work and references
44 © 2017 Nokia Public • Branching process result : − ρ < 1 : the mean population size is finite, − ρ = 1 : the mean population size is infinite. • Queueing reformulation :
− ρ < 1 : (Yn)n∈N is positive recurrent, − ρ = 1 : (Yn)n∈N is null recurrent.
Positive recurrence
• What we have shown so far :
− ρ ≤ 1 : (Yn)n∈N is recurrent. − ρ > 1 : (Yn)n∈N is transcient.
45 © 2017 Nokia Public • Queueing reformulation :
− ρ < 1 : (Yn)n∈N is positive recurrent, − ρ = 1 : (Yn)n∈N is null recurrent.
Positive recurrence
• What we have shown so far :
− ρ ≤ 1 : (Yn)n∈N is recurrent. − ρ > 1 : (Yn)n∈N is transcient.
• Branching process result : − ρ < 1 : the mean population size is finite, − ρ = 1 : the mean population size is infinite.
45 © 2017 Nokia Public Positive recurrence
• What we have shown so far :
− ρ ≤ 1 : (Yn)n∈N is recurrent. − ρ > 1 : (Yn)n∈N is transcient.
• Branching process result : − ρ < 1 : the mean population size is finite, − ρ = 1 : the mean population size is infinite. • Queueing reformulation :
− ρ < 1 : (Yn)n∈N is positive recurrent, − ρ = 1 : (Yn)n∈N is null recurrent.
45 © 2017 Nokia Public • Mean population size +∞∑ E = + E × (N) 1 k (N) pk k(=0 ) +∞∑ = + × E 1 kpk (N) k=0 • Mean number of children per node +∞∑ = ρ kpk k=0 • We obtain E(N) = 1 + ρE(N)
Mean population size
• N = total population size
N = 2 N2= 6
46 © 2017 Nokia Public • Mean number of children per node +∞∑ = ρ kpk k=0 • We obtain E(N) = 1 + ρE(N)
Mean population size
• N = total population size • Mean population size +∞∑ E = + E × (N) 1 k (N) pk k(=0 ) +∞∑ = + × E 1 kpk (N) k=0
N = 2 N2= 6
46 © 2017 Nokia Public • We obtain E(N) = 1 + ρE(N)
Mean population size
• N = total population size • Mean population size +∞∑ E = + E × (N) 1 k (N) pk k(=0 ) +∞∑ = + × E 1 kpk (N) k=0 • Mean number of children per node +∞∑ kp = ρ = k N2= 2 k=0 N 6
46 © 2017 Nokia Public Mean population size
• N = total population size • Mean population size +∞∑ E = + E × (N) 1 k (N) pk k(=0 ) +∞∑ = + × E 1 kpk (N) k=0 • Mean number of children per node +∞∑ kp = ρ = k N2= 2 k=0 N 6 • We obtain E(N) = 1 + ρE(N)
46 © 2017 Nokia Public • If ρ > 1, then E(N) > 1 + E(N) > E(N) → E(N) = +∞
• If ρ = 1, then E(N) = 1 + E(N) → E(N) = +∞
• What if ρ < 1 ? We can’t conclude ! → E < +∞ E = 1 If (N) , then (N) 1−ρ → Study the population size generation by generation
Mean population size
• We obtain E(N) = 1 + ρE(N)
N = 2 N2= 6
47 © 2017 Nokia Public • If ρ = 1, then E(N) = 1 + E(N) → E(N) = +∞
• What if ρ < 1 ? We can’t conclude ! → E < +∞ E = 1 If (N) , then (N) 1−ρ → Study the population size generation by generation
Mean population size
• We obtain E(N) = 1 + ρE(N)
• If ρ > 1, then E(N) > 1 + E(N) > E(N) → E(N) = +∞
N = 2 N2= 6
47 © 2017 Nokia Public • What if ρ < 1 ? We can’t conclude ! → E < +∞ E = 1 If (N) , then (N) 1−ρ → Study the population size generation by generation
Mean population size
• We obtain E(N) = 1 + ρE(N)
• If ρ > 1, then E(N) > 1 + E(N) > E(N) → E(N) = +∞
• If ρ = 1, then E(N) = 1 + E(N) → E(N) = +∞ N = 2 N2= 6
47 © 2017 Nokia Public Mean population size
• We obtain E(N) = 1 + ρE(N)
• If ρ > 1, then E(N) > 1 + E(N) > E(N) → E(N) = +∞
• If ρ = 1, then E(N) = 1 + E(N) → E(N) = +∞ N = 2 N2= 6 • What if ρ < 1 ? We can’t conclude ! → E < +∞ E = 1 If (N) , then (N) 1−ρ → Study the population size generation by generation
47 © 2017 Nokia Public • For each k ≥ 1, E = E E | (Nk) ( (Nk Nk−1)) = E ρ ( Nk−1) = ρE (Nk−1)
• E(N0) = 1
• E = ρk By induction, (Nk)
Mean number of nodes per generation
• = Nk number of nodes at generation k
N1 = 2
N2 = 2
N3 = 1
48 © 2017 Nokia Public • E(N0) = 1
• E = ρk By induction, (Nk)
Mean number of nodes per generation
• = Nk number of nodes at generation k
• For each k ≥ 1, E = E E | (Nk) ( (Nk Nk−1)) = E ρ ( Nk−1) = = ρE N1 2 (Nk−1)
N2 = 2
N3 = 1
48 © 2017 Nokia Public • E = ρk By induction, (Nk)
Mean number of nodes per generation
• = Nk number of nodes at generation k
• For each k ≥ 1, E = E E | (Nk) ( (Nk Nk−1)) = E ρ ( Nk−1) = = ρE N1 2 (Nk−1)
• E(N0) = 1 N2 = 2
N3 = 1
48 © 2017 Nokia Public Mean number of nodes per generation
• = Nk number of nodes at generation k
• For each k ≥ 1, E = E E | (Nk) ( (Nk Nk−1)) = E ρ ( Nk−1) = = ρE N1 2 (Nk−1)
• E(N0) = 1 N2 = 2
• E = ρk By induction, (Nk)
N3 = 1
48 © 2017 Nokia Public Hence, { +∞ if ρ ≥ 1 E(N) = 1 ρ < 1−ρ if 1
+∞∑ • = Population size : N Nk k=0 • Mean population size ( ) +∞∑ +∞∑ +∞∑ E = E = E = ρk (N) Nk (Nk) k=0 k=0 k=0
Mean population size
• E = ρk By induction, (Nk)
N2 = 2
49 © 2017 Nokia Public Hence, { +∞ if ρ ≥ 1 E(N) = 1 ρ < 1−ρ if 1
• Mean population size ( ) +∞∑ +∞∑ +∞∑ E = E = E = ρk (N) Nk (Nk) k=0 k=0 k=0
Mean population size
• E = ρk By induction, (Nk) +∞∑ • = Population size : N Nk k=0
N2 = 2
49 © 2017 Nokia Public Hence, { +∞ if ρ ≥ 1 E(N) = 1 ρ < 1−ρ if 1
Mean population size
• E = ρk By induction, (Nk) +∞∑ • = Population size : N Nk k=0 • Mean population size ( ) +∞∑ +∞∑ +∞∑ E = E = E = ρk (N) Nk (Nk) k=0 k=0 k=0
N2 = 2
49 © 2017 Nokia Public Mean population size
• E = ρk By induction, (Nk) +∞∑ • = Population size : N Nk k=0 • Mean population size ( ) +∞∑ +∞∑ +∞∑ E = E = E = ρk (N) Nk (Nk) k=0 k=0 k=0 Hence, N2 = 2 { +∞ if ρ ≥ 1 E(N) = 1 ρ < 1−ρ if 1
49 © 2017 Nokia Public By the same reasonning, we obtain { +∞ if ρ ≥ 1 E(B) = 1 1 ρ < µ 1−ρ if 1
• Mean length of the busy period 1 +∞∑ E(B) = + kE(B) × p µ k k(=0 ) 1 +∞∑ = + kp × E(B) µ k k=0 1 = + ρE(B) µ
Length of a busy period
• B = time length of the busy period
N2 = 2
50 © 2017 Nokia Public By the same reasonning, we obtain { +∞ if ρ ≥ 1 E(B) = 1 1 ρ < µ 1−ρ if 1
Length of a busy period
• B = time length of the busy period • Mean length of the busy period 1 +∞∑ E(B) = + kE(B) × p µ k k(=0 ) 1 +∞∑ = + kp × E(B) µ k k=0 1 = + ρE(B) µ N2 = 2
50 © 2017 Nokia Public Length of a busy period
• B = time length of the busy period • Mean length of the busy period 1 +∞∑ E(B) = + kE(B) × p µ k k(=0 ) 1 +∞∑ = + kp × E(B) µ k k=0 1 = + ρE(B) µ N = 2 By the same reasonning, we obtain 2 { +∞ if ρ ≥ 1 E(B) = 1 1 ρ < µ 1−ρ if 1
50 © 2017 Nokia Public Length of a busy period µ = 1 10 1 1−ρ 8
6 Time 4
2
0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Load ρ
51 © 2017 Nokia Public M/G/1 queue
Branching process
Stability Recurrence and transcience Positive recurrence
Related work and references
52 © 2017 Nokia Public Related work
• M/G/1 queue with LCFS service discipline • GI/M/1 queue • Pooling systems
53 © 2017 Nokia Public Bibliography David G. Kendall (1951). “Some Problems in the Theory of Queues”. In : Journal of the Royal Statistical Society. Series B (Methodological) 13.2, p. 151–185. David G. Kendall (1953). “Stochastic Processes Occurring in the Theory of Queues and their Analysis by the Method of the Imbedded Markov Chain”. In : The Annals of Mathematical Statistics 24.3, p. 338–354. Pierre Brémaud (2013). Markov Chains : Gibbs Fields, Monte Carlo Simulation, and Queues. Texts in Applied Mathematics. Springer New York. Berestycki Nathanaël (2014). Applied Probability - Part II. Course. University of Cambridge. Yoram Clapper (2017). Branching Processes in Queuing Theory. Bachelor’s thesis mathematics. University of Amsterdam.
54 © 2017 Nokia Public