IS 709/809: Computational Methods in IS Research. Simple Markovian Queueing Model

Similar documents
Analysis of System Performance IN2072 Chapter 5 Analysis of Non Markov Systems

EP2200 Queueing theory and teletraffic systems. Queueing networks. Viktoria Fodor KTH EES/LCN KTH EES/LCN

Channel Models with Memory. Channel Models with Memory. Channel Models with Memory. Channel Models with Memory

Outline. Basic Components of a Queue. Queueing Notation. EEC 686/785 Modeling & Performance Evaluation of Computer Systems.

Simulation Output Analysis

Random Variables. ECE 313 Probability with Engineering Applications Lecture 8 Professor Ravi K. Iyer University of Illinois

Some queue models with different service rates. Július REBO, Žilinská univerzita, DP Prievidza

Markov chain. Markov Chain? Markov chain. Discrete-Time Markov Chain (DTMC) Memoryless (Markov) Property: Given the present state, Markov Chain

Chain Rules for Entropy

Summary of the lecture in Biostatistics

Functions of Random Variables

APPENDIX A: ELEMENTS OF QUEUEING THEORY

2. Independence and Bernoulli Trials

Optimal Strategy Analysis of an N-policy M/E k /1 Queueing System with Server Breakdowns and Multiple Vacations

Special Instructions / Useful Data

THE ROYAL STATISTICAL SOCIETY GRADUATE DIPLOMA

Lecture 9. Some Useful Discrete Distributions. Some Useful Discrete Distributions. The observations generated by different experiments have

Likewise, properties of the optimal policy for equipment replacement & maintenance problems can be used to reduce the computation.

Application of Generating Functions to the Theory of Success Runs

å 1 13 Practice Final Examination Solutions - = CS109 Dec 5, 2018

A Markov Chain Competition Model

Continuous Distributions

Random Variables and Probability Distributions

CHAPTER 6. d. With success = observation greater than 10, x = # of successes = 4, and

Parameter, Statistic and Random Samples

Estimation of Stress- Strength Reliability model using finite mixture of exponential distributions

Chapter 4 (Part 1): Non-Parametric Classification (Sections ) Pattern Classification 4.3) Announcements

Chapter 5 Properties of a Random Sample

D KL (P Q) := p i ln p i q i

Introduction to local (nonparametric) density estimation. methods

Birth Death Processes

{ }{ ( )} (, ) = ( ) ( ) ( ) Chapter 14 Exercises in Sampling Theory. Exercise 1 (Simple random sampling): Solution:

Chapter 3 Sampling For Proportions and Percentages

The Mathematical Appendix

Chapter 14 Logistic Regression Models

. The set of these sums. be a partition of [ ab, ]. Consider the sum f( x) f( x 1)

STK3100 and STK4100 Autumn 2017

On the introductory notes on Artin s Conjecture

Study of Impact of Negative Arrivals in Single. Server Fixed Batch Service Queueing System. with Multiple Vacations

STRONG CONSISTENCY FOR SIMPLE LINEAR EV MODEL WITH v/ -MIXING

Random Variate Generation ENM 307 SIMULATION. Anadolu Üniversitesi, Endüstri Mühendisliği Bölümü. Yrd. Doç. Dr. Gürkan ÖZTÜRK.

Dr. Shalabh Department of Mathematics and Statistics Indian Institute of Technology Kanpur

LINEAR REGRESSION ANALYSIS

Introduction to Probability

Unsupervised Learning and Other Neural Networks

PTAS for Bin-Packing

Discrete Mathematics and Probability Theory Fall 2016 Seshia and Walrand DIS 10b

Bounds on the expected entropy and KL-divergence of sampled multinomial distributions. Brandon C. Roy

ANALYSIS ON THE NATURE OF THE BASIC EQUATIONS IN SYNERGETIC INTER-REPRESENTATION NETWORK

Lecture 3 Probability review (cont d)

Introduction to Econometrics (3 rd Updated Edition, Global Edition) Solutions to Odd-Numbered End-of-Chapter Exercises: Chapter 9

Continuous Random Variables: Conditioning, Expectation and Independence

ON BIVARIATE GEOMETRIC DISTRIBUTION. K. Jayakumar, D.A. Mundassery 1. INTRODUCTION

8. Queueing systems lect08.ppt S Introduction to Teletraffic Theory - Fall

Chapter 5 Properties of a Random Sample

2SLS Estimates ECON In this case, begin with the assumption that E[ i

CS286.2 Lecture 4: Dinur s Proof of the PCP Theorem

2.28 The Wall Street Journal is probably referring to the average number of cubes used per glass measured for some population that they have chosen.

Lecture 07: Poles and Zeros

Departure Process from a M/M/m/ Queue

Pr[X (p + t)n] e D KL(p+t p)n.

Open and Closed Networks of M/M/m Type Queues (Jackson s Theorem for Open and Closed Networks) Copyright 2015, Sanjay K. Bose 1

X ε ) = 0, or equivalently, lim

Integral Generalized Binomial Coefficients of Multiplicative Functions

Analysis of Variance with Weibull Data

STK3100 and STK4100 Autumn 2018

D. VQ WITH 1ST-ORDER LOSSLESS CODING

Class 13,14 June 17, 19, 2015

MATH 247/Winter Notes on the adjoint and on normal operators.

CHAPTER VI Statistical Analysis of Experimental Data

UNIVERSITY OF OSLO DEPARTMENT OF ECONOMICS

Entropy, Relative Entropy and Mutual Information

THE ROYAL STATISTICAL SOCIETY HIGHER CERTIFICATE

Assignment 5/MATH 247/Winter Due: Friday, February 19 in class (!) (answers will be posted right after class)

Lecture 3. Sampling, sampling distributions, and parameter estimation

Complete Convergence and Some Maximal Inequalities for Weighted Sums of Random Variables

Ordinary Least Squares Regression. Simple Regression. Algebra and Assumptions.

Arithmetic Mean and Geometric Mean

The E vs k diagrams are in general a function of the k -space direction in a crystal

THE ROYAL STATISTICAL SOCIETY GRADUATE DIPLOMA

Homework 1: Solutions Sid Banerjee Problem 1: (Practice with Asymptotic Notation) ORIE 4520: Stochastics at Scale Fall 2015

Steady-state Behavior of a Multi-phase M/M/1 Queue in Random Evolution subject to Catastrophe failure

Section l h l Stem=Tens. 8l Leaf=Ones. 8h l 03. 9h 58

Periodic Table of Elements. EE105 - Spring 2007 Microelectronic Devices and Circuits. The Diamond Structure. Electronic Properties of Silicon

Dr. Shalabh. Indian Institute of Technology Kanpur

Chapter 9 Jordan Block Matrices

Minimizing Total Completion Time in a Flow-shop Scheduling Problems with a Single Server

Econometric Methods. Review of Estimation

UNIVERSITY OF OSLO DEPARTMENT OF ECONOMICS

Two Fuzzy Probability Measures

Optimum Probability Distribution for Minimum Redundancy of Source Coding

ρ < 1 be five real numbers. The

Chapter 4 Multiple Random Variables

STA 105-M BASIC STATISTICS (This is a multiple choice paper.)

Waiting Time Distribution of Demand Requiring Multiple Items under a Base Stock Policy

Unit 9. The Tangent Bundle

Block-Based Compact Thermal Modeling of Semiconductor Integrated Circuits

Bayes (Naïve or not) Classifiers: Generative Approach

Analysis of a Repairable (n-1)-out-of-n: G System with Failure and Repair Times Arbitrarily Distributed

STA 108 Applied Linear Models: Regression Analysis Spring Solution for Homework #1

Transcription:

IS 79/89: Comutatoal Methods IS Research Smle Marova Queueg Model Nrmalya Roy Deartmet of Iformato Systems Uversty of Marylad Baltmore Couty www.umbc.edu

Queueg Theory Software QtsPlus software The software s avalable from the Wley ublc ft server ft://ft.wley.com/ublc/sc_tech_med/queueg_theory Based o the lectures we wll smulate some queueg theory models usg ths software QtsPlus cotas models the followg areas: Fte ad fte brth-death rocess models Sgle server queueg models Multle server queueg models Bul queueg models Prorty queueg models Networ queueg models

Itroducto Develo a broad class of smle queueg models theory of Brth-Death rocess A Brth-Death rocess s a secfc tye of cotuoustme Marov cha Examle of queues that ca be modeled are M/M/, M/M/c, M/M/c/K, M/M/c/c, M/M/ etc Cossts of a set of states {,,,..} deotg the oulato of some system State trasto occur as ut jums u or dow from the curret state

Brth-Death Process (BDP) Secal case of Marov cha wth restrcted trastos Rate trasto dagram for a brth-death rocess Trastos are restrcted to eghborg states oly (rocess state ca chage oly to state + or -)

Brth-Death Processes System s state The tme utl the ext arrval ( brth ) s a exoetal radom varable wth rate λ At a arrval system moves from state to state (+) The tme utl the ext dearture ( death ) s a exoetal radom varable wth rate μ At a dearture, the system moves from state to state (-) deotes the log-term fracto of tme the system s state Ths s a cotuous tme Marov cha States deote the umber of customers the system Brths corresod to customer arrvals Deaths corresod to customer deartures

Brth-Death Processes Cocet of flow balace I steady state rate of trasto out of a gve state = rate of trasto to that state LHS deotes the rate of trasto out of state RHS deotes the rate of trasto to state - + ( ) ( )

Brth-Death Processes - + Trasto out of state ca be ether uward or dowward ( ) s the log-term rate of trasto out of state Trasto to state ca occur ether from the state below (-) or the state above (+) s the log-term rate of trasto to state reresets a flow balace for the boudary state

...... Equlbrum Soluto of BDP ) ( 3 3 3 3 Prove that

Use mathematcal ducto to rove that Equlbrum Soluto of BDP...

Posso Process Secal case of brth-death rocesses Posso rocess s a coutg rocess (ure-brth rocess) havg costat deedet ad statoary cremets (costat brth rate)

Suerosto of Posso Processes Aggregatg multle Posso rocesses roduces a hgher rate Posso rocess (whose rate s sum of the rate of dvdual Posso rocesses)

Decomosto of Posso Processes Decomosg a Posso rocess stream to multle stream results multle Posso streams of lower-rate.

Stablty of a Brth-Death rocess We desre that queues do t grow ubouded Posso rocess (ustable sce the death rate s zero; sze grows ubouded) For a BDP to be stable, there must exst some state j such that μ j s greater tha λ j for all j > j

Sgle Server Queues

M/M/ queue Iterarrval tmes ad servce tmes are exoetally dstrbuted Probablty desty fuctos a( t) t e b( t) The rate of arrvals λ s fxed, regardless of the umber the system The rate of the server μ s fxed, regardless of the umber the system t e M/M/ s a brth-death rocess wth ad ( ) ( )

M/M/ queue (a smlfed BDP) M/M/ queue s a brth-death rocess wth λ = λ ad μ = μ Rate trasto dagram for the M/M/ queue The flow-balace equatos for ths system are ) ( ) ( Alteratvely, these ca be wrtte as ) (

M/M/ queue Soluto Solve for { } usg a teratve method Iteratvely use the balace equatos to obta a sequece of state robabltes,, 3,.., each terms of Verfy that cojecture s correct usg mathematcal ducto M/M/ system s a brth-death rocess wth costat brth ad death rates, we ca drectly aly wth ad The soluto of M/M/ system s

M/M/ queue Soluto Probabltes of { } must sum to for a sgle server queue where ρ s the traffc testy or utlzato Ths s cosstet wth geeral result for for all G/G/ queues G/G/ emty system robablty

M/M/ queue Steady State Soluto The full steady state soluto for M/M/ system s the geometrc robablty fucto ( ) ( ) Note the exstece of a steady state soluto deeds o the codto that Equvaletly Itutvely f the mea arrval rate > mea servce rate Server gets further ad further behd; system sze creases wthout boud over tme Why there s o steady state soluto whe

M/M/ queue Steady State Soluto Why there s o steady state soluto whe Ifte buld u As the queue grows t s more ad more dffcult for the server to decrease the queue Average servce rate s o hgher tha the average arrval rate

M/M/ Examle Cars arrve at a solated toll booth as Posso rocess wth arrval rate λ=.6 cars er mute. The servce requred by a customer s a exoetal radom varable wth exected value of /μ=.3 mutes. What are the lmtg state rob. for N, umber of cars at toll booth? Sce ths s a M/M/ queue (Posso arrvals/ deartures; sgle server) wth ρ=λ/μ=.8, the lmtg state robabltes are: = (-ρ) ρ =(.8)(.8), =,,. What s the robablty that the toll booth has zero cars steady-state? =.8

M/M/ Examle Cosder a l wth Posso arrvals (rate of 3 acets/sec) such that the tme tae to servce a acet s exoetally dstrbuted. Assume mea acet legth of 5 bytes ad l caacty of.5 Mbs. Probablty that the l s queue has, ad acets resectvely? Arrval rate λ = 3 acets/secod. L caacty =.5 Mbs Rate of servg acets μ =.5 Mbs/ (5 8) bts = 375 acets/s Traffc testy or utlzato ρ=λ\μ =.8 Probablty of oe acet queue = ρ (-ρ) =.8 (.) =.8 (ote: oe acet queue meas two acets system) Smlarly, 3 =., ad =.7

Measures of Effectveess Steady state robablty dstrbuto for the system sze Calculate the system s measure of effectveess Exected umber the system at steady state Exected umber the queue at steady state N reresets the radom varable umber of customers the system steady state L reresets ts exected value L E[ N] ( )

3...) 3 (... 3 ) ( d d d d Exected umber the system for M/M/ at steady state s L ) ( ) ( ) ( ) ( L Exected Number the System (M/M/)

Exected Number the Queue (M/M/) Number of customers queue steady state = N q L q reresets ts exected value For all sgle-chael, oe-at-a-tme servce queues Mea queue legth ) ( ) ( ] [ L N E L q q ) ( L L q ) ( L q

Exected Watg Tme Exected steady state system watg tme W ad le delay W q Use Lttle s Theorem of L ad L q ) ( L W ) ( q q L W

Summary Brth-death rocess Steady state robabltes ad erformace measures M/M/ ad M/M/C (Erlag C formula) ad M/M/ Fte buffer: M/M/c/K ad M/M//K ad M/M/c/c (Erlag s Loss formula/erlag B formula) M/M/c ad M/M/ are very useful models for multle servers queueg systems Fte buffered M/M//K ad M/M/c/c rovde useful egeerg rovsog gudeles

Questos?