This preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: IEOR 4703: Homework 9 1. Gibbs sampler for a closed Jackson queueing network: (READ LECTURE NOTES 10 FOR REFERENCE HERE; SECTION 2.2) Consider a closed queueing network with c = 10 nodes (singleserver FIFO queues), and M = 50 customers, in which the 10 × 10 transition matrix P for customer routing is known to be irreducible and doubly stochastic , so that the unique probability solution to λ = λP is the discrete uniform distribution over the set { 1 , 2 ,..., 10 } ; λ i = 1 / 10 , i ∈ { 1 , 2 ,..., 10 } . Further assume that the service times at node i are iid exponentially distributed with rate μ i = 1 /i, i ∈ { 1 , 2 ,..., 10 } . For then ρ i = λ i /μ i = i/ 10, and the largest one is ρ c = ρ 10 = 1. As a result, the stationary distribution π of ( X 1 ( t ) ,...,X 9 ( t )) is of the form π ( x ) = Kg ( x ) def = K 9 Y i =1 ρ i x i , x ∈ S , (1) where S = { x ∈ N 9 + : 9 X i =1 x i ≤ M } , and K 1 = X x ∈S g ( x ) , (a constant for which we typically do not know, and will not need in what follows).(a constant for which we typically do not know, and will not need in what follows)....
View
Full Document
 Spring '07
 sigman
 Probability theory, Gibbs sampler, j=i xj x=0, singleserver FIFO queues, closed Jackson queueing

Click to edit the document details