Stochastic

# s depending on the number of servers that are busy

This preview shows page 1. Sign up to view the full content.

This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: C = ¯ y 2A ⇡ (y ). Then ⌫ (x) = ⇡ (x)/C is a stationary distribution for Yt . 139 4.5. MARKOVIAN QUEUES Proof. If x, y 2 A then detailed balance for Xt implies ⇡ (x)q (x, y ) = ⇡ (y )q (y, x). From this it follows that ⌫ (x)¯(x, y ) = ⌫ (y )¯(y, x) so ⌫ satisﬁes the detailed q q balance condition for Yt . It follows from Lemma 4.7 that ✓ ◆n ⇡ (n) = /C µ for 1 n N To compute the normalizing constant, we recall that if ✓ 6= 1, then N X ✓n = n=0 Suppose now that 1 ✓N +1 1✓ (4.24) 6= µ. Using (4.24), we see that C= 1 ( /µ)N +1 1 /µ so the stationary distribution is given by ✓ ◆n 1 /µ ⇡ (n) = 1 ( /µ)N +1 µ for 0 n N (4.25) The new formula is similar to the old one in (4.23) and when < µ reduces to it as N ! 1. Of course, when the waiting room is ﬁnite, the state space is ﬁnite and we always have a stationary distribution, even when > µ. The analysis above has been restricted to 6= µ. However, it is easy to see that when = µ the stationary distribution is ⇡ (n) = 1/(N + 1) for 0 n N . To check formula (4.25), we note that the barbershop chain, Example 4.14, has this form with N = 3,...
View Full Document

## This document was uploaded on 03/06/2014 for the course MATH 4740 at Cornell.

Ask a homework question - tutors are online