Top Banner
CDA6530: Performance Models of Computers and Networks Chapter 7: Basic Queuing Networks
14

CDA6530: Performance Models of Computers and Networks Chapter 7: Basic Queuing Networks

Feb 05, 2016

Download

Documents

palila

CDA6530: Performance Models of Computers and Networks Chapter 7: Basic Queuing Networks. TexPoint fonts used in EMF. Read the TexPoint manual before you delete this box.: A A A A A A A A A A A A. Open Queuing Network. Jobs arrive from external sources, circulate, and eventually depart. - PowerPoint PPT Presentation
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Page 1: CDA6530: Performance Models of Computers and Networks Chapter 7: Basic Queuing Networks

CDA6530: Performance Models of Computers and Networks

Chapter 7: Basic Queuing Networks

Page 2: CDA6530: Performance Models of Computers and Networks Chapter 7: Basic Queuing Networks

2

Open Queuing Network

Jobs arrive from external sources, circulate, and eventually depart

Page 3: CDA6530: Performance Models of Computers and Networks Chapter 7: Basic Queuing Networks

3

Closed Queuing Network

Fixed population of K jobs circulate continuously and never leave Previous machine-repairman problem

Page 4: CDA6530: Performance Models of Computers and Networks Chapter 7: Basic Queuing Networks

4

Feed-Forward QNs

Consider two queue tandem system

Q: how to model? System is a continuous-time Markov chain (CTMC) State (N1(t), N2(t)), assume to be stable ¼(i,j) =P(N1=i, N2=j) Draw the state transition diagram

But what is the arrival process to the second queue?

Page 5: CDA6530: Performance Models of Computers and Networks Chapter 7: Basic Queuing Networks

5

Poisson in ) Poisson out

Burke’s Theorem: Departure process of M/M/1 queue is Poisson with rate λ independent of arrival process.

Poisson process addition, thinning Two independent Poisson arrival processes adding

together is still a Poisson (¸=¸1+¸2) For a Poisson arrival process, if each customer lefts

with prob. p, the remaining arrival process is still a Poisson (¸ = ¸1¢ p)

Why?

Page 6: CDA6530: Performance Models of Computers and Networks Chapter 7: Basic Queuing Networks

6

State transition diagram: (N1, N2), Ni=0,1,2,

Page 7: CDA6530: Performance Models of Computers and Networks Chapter 7: Basic Queuing Networks

7

For a k queue tandem system with Poisson arrival and expo. service time

Jackson’s theorem:

Above formula is true when there are feedbacks among different queues Each queue behaves as M/M/1 queue in isolation

Page 8: CDA6530: Performance Models of Computers and Networks Chapter 7: Basic Queuing Networks

8

Example

¸i: arrival rate at queue i

Why?

Why?

In M/M/1:

Page 9: CDA6530: Performance Models of Computers and Networks Chapter 7: Basic Queuing Networks

9

T(i): response time for a job enters queue i

Why?

In M/M/1:

E [T (1)] = 1=(¹ 1 ¡ ¸1) + E [T (2)]=2E [T (2)] = 1=(¹ 2 ¡ ¸2) + E [T (1)]=4

Page 10: CDA6530: Performance Models of Computers and Networks Chapter 7: Basic Queuing Networks

10

Extension

results hold when nodes are multiple server nodes (M/M/c), infinite server nodes finite buffer nodes (M/M/c/K) (careful about interpretation of results), PS (process sharing) single server with arbitrary service time distr.

Page 11: CDA6530: Performance Models of Computers and Networks Chapter 7: Basic Queuing Networks

11

Closed QNs

Fixed population of N jobs circulating among M queues. single server at each queue, exponential service

times, mean 1/μi for queue i routing probabilities pi,j, 1 ≤ i, j ≤ M visit ratios, {vi}. If v1 = 1, then vi is mean number of

visits to queue i between visits to queue 1

: throughput of queue i,°i

Page 12: CDA6530: Performance Models of Computers and Networks Chapter 7: Basic Queuing Networks

12

Example

Open QN has infinite no. of states Closed QN is simpler

How to define states? No. of jobs in each queue

Page 13: CDA6530: Performance Models of Computers and Networks Chapter 7: Basic Queuing Networks

13

Page 14: CDA6530: Performance Models of Computers and Networks Chapter 7: Basic Queuing Networks

14

Steady State Solution

Theorem (Gordon and Newell)

For previous example when p1=0.75 , vi?