Analysis of Discrete Time Queues (Section 4.6)

Similar documents
Equilibrium Analysis of the M/G/1 Queue

Departure Process from a M/M/m/ Queue

Introduction to Continuous-Time Markov Chains and Queueing Theory

TCOM 501: Networking Theory & Fundamentals. Lecture 7 February 25, 2003 Prof. Yannis A. Korilis

CS 798: Homework Assignment 2 (Probability)

On the relationships among queue lengths at arrival, departure, and random epochs in the discrete-time queue with D-BMAP arrivals

Order Full Rate, Leadtime Variability, and Advance Demand Information in an Assemble- To-Order System

Markov chains. Definition of a CTMC: [2, page 381] is a continuous time, discrete value random process such that for an infinitesimal

Application of Queuing Theory to Waiting Time of Out-Patients in Hospitals.

Queueing Networks II Network Performance

SOJOURN TIME IN A QUEUE WITH CLUSTERED PERIODIC ARRIVALS

6. Stochastic processes (2)

6. Stochastic processes (2)

Applied Stochastic Processes

COMPLETE BUFFER SHARING IN ATM NETWORKS UNDER BURSTY ARRIVALS

CS-433: Simulation and Modeling Modeling and Probability Review

Adaptive Dynamical Polling in Wireless Networks

3.1 Expectation of Functions of Several Random Variables. )' be a k-dimensional discrete or continuous random vector, with joint PMF p (, E X E X1 E X

arxiv: v1 [math.pr] 1 Aug 2014

Suggested solutions for the exam in SF2863 Systems Engineering. June 12,

Distributions /06. G.Serazzi 05/06 Dimensionamento degli Impianti Informatici distrib - 1

12. The Hamilton-Jacobi Equation Michael Fowler

CHAPTER 5 NUMERICAL EVALUATION OF DYNAMIC RESPONSE

Lecture 3: Probability Distributions

Eigenvalues of Random Graphs

Dynamic Programming. Lecture 13 (5/31/2017)

The optimal delay of the second test is therefore approximately 210 hours earlier than =2.

EPR Paradox and the Physical Meaning of an Experiment in Quantum Mechanics. Vesselin C. Noninski

Response time in a tandem queue with blocking, Markovian arrivals and phase-type services

Markov Chain Monte Carlo (MCMC), Gibbs Sampling, Metropolis Algorithms, and Simulated Annealing Bioinformatics Course Supplement

Probability and Random Variable Primer

See Book Chapter 11 2 nd Edition (Chapter 10 1 st Edition)

The Expectation-Maximization Algorithm

Review: Discrete Event Random Processes. Hongwei Zhang

APPROXIMATE PRICES OF BASKET AND ASIAN OPTIONS DUPONT OLIVIER. Premia 14

Queuing system theory

Chapter 20 Duration Analysis

PHYS 705: Classical Mechanics. Calculus of Variations II

A note on almost sure behavior of randomly weighted sums of φ-mixing random variables with φ-mixing weights

Engineering Risk Benefit Analysis

Chapter 12. Ordinary Differential Equation Boundary Value (BV) Problems

ECE697AA Lecture 17. Birth-Death Processes

Problem Set 9 - Solutions Due: April 27, 2005

Learning from Data 1 Naive Bayes

10. Canonical Transformations Michael Fowler

MASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.265/15.070J Fall 2013 Lecture 12 10/21/2013. Martingale Concentration Inequalities and Applications

x = , so that calculated

Convergence of random processes

Notes prepared by Prof Mrs) M.J. Gholba Class M.Sc Part(I) Information Technology

χ x B E (c) Figure 2.1.1: (a) a material particle in a body, (b) a place in space, (c) a configuration of the body

Computation of Higher Order Moments from Two Multinomial Overdispersion Likelihood Models

Statistics and Probability Theory in Civil, Surveying and Environmental Engineering

UNIVERSITY OF TORONTO Faculty of Arts and Science. December 2005 Examinations STA437H1F/STA1005HF. Duration - 3 hours

Analysis of Queuing Delay in Multimedia Gateway Call Routing

LINEAR REGRESSION ANALYSIS. MODULE VIII Lecture Indicator Variables

ANSWERS. Problem 1. and the moment generating function (mgf) by. defined for any real t. Use this to show that E( U) var( U)

NUMERICAL DIFFERENTIATION

18.1 Introduction and Recap

Vapnik-Chervonenkis theory

Lecture 2 M/G/1 queues. M/G/1-queue

Econ107 Applied Econometrics Topic 3: Classical Model (Studenmund, Chapter 4)

C/CS/Phy191 Problem Set 3 Solutions Out: Oct 1, 2008., where ( 00. ), so the overall state of the system is ) ( ( ( ( 00 ± 11 ), Φ ± = 1

Lecture 10 Support Vector Machines II

SELECTED PROOFS. DeMorgan s formulas: The first one is clear from Venn diagram, or the following truth table:

Chapter 13: Multiple Regression

P exp(tx) = 1 + t 2k M 2k. k N

Lecture 4 Hypothesis Testing

Lecture 4: November 17, Part 1 Single Buffer Management

ANSWERS CHAPTER 9. TIO 9.2: If the values are the same, the difference is 0, therefore the null hypothesis cannot be rejected.

6.3.4 Modified Euler s method of integration

Transient results for M/M/1/c queues via path counting. M. Hlynka*, L.M. Hurajt and M. Cylwa

Randomness and Computation

Notes on Frequency Estimation in Data Streams

Lectures - Week 4 Matrix norms, Conditioning, Vector Spaces, Linear Independence, Spanning sets and Basis, Null space and Range of a Matrix

STAT 511 FINAL EXAM NAME Spring 2001

AP Physics 1 & 2 Summer Assignment

NP-Completeness : Proofs

Rapporto di Ricerca CS Simonetta Balsamo, Andrea Marin. On representing multiclass M/M/k queue by Generalized Stochastic Petri Net

ELASTIC WAVE PROPAGATION IN A CONTINUOUS MEDIUM

EURANDOM PREPRINT SERIES October, Performance analysis of polling systems with retrials and glue periods

Elevator-Type Polling Systems

Physics 181. Particle Systems

Markov Chain Monte Carlo Lecture 6

Chapter 2 Transformations and Expectations. , and define f

Meenu Gupta, Man Singh & Deepak Gupta

Case Study of Markov Chains Ray-Knight Compactification

The Multiple Classical Linear Regression Model (CLRM): Specification and Assumptions. 1. Introduction

Module 1 : The equation of continuity. Lecture 1: Equation of Continuity

Bounds on the bias terms for the Markov reward approach

A Bayesian Approach to Arrival Rate Forecasting for Inhomogeneous Poisson Processes for Mobile Calls

Stanford University CS254: Computational Complexity Notes 7 Luca Trevisan January 29, Notes for Lecture 7

Nice plotting of proteins II

Expected Value and Variance

Turbulence classification of load data by the frequency and severity of wind gusts. Oscar Moñux, DEWI GmbH Kevin Bleibler, DEWI GmbH

Case A. P k = Ni ( 2L i k 1 ) + (# big cells) 10d 2 P k.

Continuous Time Markov Chains

Machine learning: Density estimation

= z 20 z n. (k 20) + 4 z k = 4

P R. Lecture 4. Theory and Applications of Pattern Recognition. Dept. of Electrical and Computer Engineering /

Digital Signal Processing

Transcription:

Analyss of Dscrete Tme Queues (Secton 4.6) Copyrght 2002, Sanjay K. Bose Tme axs dvded nto slots slot slot boundares Arrvals can only occur at slot boundares Servce to a job can only start at a slot boundary Servce duraton s always an ntegral multple of the slot duraton Copyrght 2002, Sanjay K. Bose 2

Assume that the number of jobs that arrve n successve slots are ndependent, dentcally dstrbuted (..d.) random varables Geo Arrval Process: Geometrc (or Bernoull) Arrval Process Only one job can arrve n a slot wth probablty λ and that no jobs arrve n a slot wth probablty -λ, (0<λ<) The nter-arrval tme I s geometrcally dstrbuted wth wth mean /λ and wth probablty- P{I k slots} λ(-λ) -k for k,2,... λ( Generatng Functon of the number of arrvals n a slot Copyrght 2002, Sanjay K. Bose 3 Geo [X] Arrval Process Batch Geometrc Arrval Process Λnumber of arrvals n a slot (d random varables) λ(k) P{Λ k arrvals n a slot} k0,,2,.. k 0 λ( k) z k Generatng Functon of the number of arrvals n a slot The followng defntons of means and factoral moments would be useful later λ E{ Λ} Λ λ λ ( ) E{ Λ )} Λ E{ Λ )...( Λ + )} Λ ( ) 3, 4,... Copyrght 2002, Sanjay K. Bose 4 2

Choce of Arrval Models Arrval Models Late Arrval Early Arrval Queue sze measured mmedately after the slot boundares Watng tme n queue wll dffer for the late and early models Copyrght 2002, Sanjay K. Bose 5 Late Arrval Model Servce Completon n n+ n+2 Arrval Tme Arrvals come late n the slot,.e. just before the slot boundary and before the servce completons due to occur at the end of that slot Number left behnd n the queue as seen by a departure wll nclude the arrvals as shown above Tme spent watng n queue Number of slots spent watng for servce not ncludng the slot n whch the job arrves For example, watng tme s zero f a job arrves at the end of the n th slot and starts servce from the begnnng of the (n+) th slot,.e. the next slot after ts arrval. Copyrght 2002, Sanjay K. Bose 6 3

Early Arrval Model Servce Completon n n+ n+2 Arrval Tme Arrvals come early n the slot,.e. just after the slot boundary. Note that servce completons occur just before the slot boundary as n the late arrval model Number left behnd n the queue as seen by a departure wll not nclude the arrvals as shown above Tme spent watng n queue Number of slots spent watng for servce ncludng the slot n whch the job arrves For example, watng tme s zero f a job arrves at the begnnng of the (n+) th slot and starts servce from that slot tself. Copyrght 2002, Sanjay K. Bose 7 Important ponts to note - Queue sze at servce completon n the early arrval model wll always be lower than n the correspondng late arrval model. Lower by the number of jobs whch actually arrve at that slot boundary The watng tme n queue wll be the same regardless of the actual model (early/late) beng used Copyrght 2002, Sanjay K. Bose 8 4

Servce Tmes Multples of slot duratons - starts and ends at the slot boundares General IID Dstrbuton for the number of slots (say random varable X) requred to complete servce to a job. Probablty Dstrbuton of Servce Tmes (n terms of the number of slots requred for servce Probablty Dstrbuton Generatng Functon Moments b ( k) P{ X k} for k,2,. k E{ X E{ X b( k) z b E{ X } B b b ( ) 2 } k } B + B 3, 4,... Copyrght 2002, Sanjay K. Bose 9 The Geo/G/ Queue (dscrete tme verson of the M/G/ queue) Late Arrval Model Separate analyss requred for the late arrval and the early arrval model Early Arrval Model Stablty requres that the traffc offered ρ λb< Copyrght 2002, Sanjay K. Bose 0 5

The Geo/G/ Queue (Late Arrval Model) n number of jobs n the queue mmedately after the servce completon of the th job a number of jobs arrvng durng the servce tme of the th job The random varables a,2,... are ndependent and dentcally dstrbuted random varables wth the generatng functon A( and mean ρ. Homogenous, Dscrete- Tme Markov Chan n + a + n + a + n 0 n (4.9) Copyrght 2002, Sanjay K. Bose At equlbrum, the steady-state dstrbuton p k of ths Markov Chan may be found as - p k lm P{ n k} k 0,,2,... and k 0 k p k z Show that, usng (4.9) under equlbrum condtons gves - ( ρ)( A( A( z wth p 0 ρ ρ λb Copyrght 2002, Sanjay K. Bose 2 6

Show that A( λ + λ leadng to ( ρ)( λ + λ λ + λ z (4.97) Derved for the nstants of servce completons of jobs Also holds for Arrval nstants Just after arbtrary slot boundary Arbtrary tme nstant on the contnuous tme axs BASTA/GASTA: Bernoull/Geometrc Arrvals See Tme Averages holds as the dscrete tme verson of PASTA Copyrght 2002, Sanjay K. Bose 3 Usng from (4.97), show that Mean number n system 2 b λ λρ N + ρ (4.98) 2( ρ) The mean tme W (n number of slots) spent by a job n the system may be found from (4.98) usng Lttle s Result NλW We can then use W q W - b and N q λw q to get the performance measures W q and N q Copyrght 2002, Sanjay K. Bose 4 7

The dstrbuton G W ( (.e. the generatng functon) of the number of slots for whch a job stays n the system may also be obtaned for a FCFS Geo/G/ queue. Usng an approach smlar to that followed for the M/G/ queue, show that for the FCFS Geo/G/ queue, we wll have - GW ( λ + λ whch leads to and G W G Wq ( ρ)( ( ( λ( ) ( ρ)( ( ( λ( ) (4.02) (4.04) followng usual ndependence argument between servce tme and tme spent watng n queue Copyrght 2002, Sanjay K. Bose 5 The Geo/G/ Queue (Early Arrval Model) n number of jobs n the queue after the servce completon of the th job and before the next possble arrval pont a number of jobs arrvng durng the servce tme of the th job a ~ number of jobs arrvng n the servce tme of the the th job mnus one slot Homogenous, Dscrete- Tme Markov Chan n + a~ + n + a + n 0 n (4.06) Copyrght 2002, Sanjay K. Bose 6 8

Generatng Functon A( of a obtaned as before Show that A( λ + λ B z A ~ ( λ + λ ) ( λ + λz From (4.06), show that ~ ( ρ)[ A( za( ] ( λ)[ A( z] and hence ( ρ)( λ + λ ( λ + λ[ λ + λ z] Copyrght 2002, Sanjay K. Bose 7 Show that, n ths case GW ( λ + λ ( λ + λ whch leads to G W ( ρ)( ( ( λ( ) Note that even though s dfferent for the late/early arrval models, G W ( s the same as clamed earler Copyrght 2002, Sanjay K. Bose 8 9

The Geo [X] /G/ Queue Summary of results gven here. Please see Sec. 4.6.2 for detals Generatng Functon of the number of jobs arrvng n a slot Generatng Functon of the number of slot requred for servce by a job Copyrght 2002, Sanjay K. Bose 9 The Geo [X} /G/ Queue (Late Arrval Model) Generatng Fn. of the number n the system just after a servce completon ( ρ)[ ] ) λ[ ) z] Generatng Fn. of the number n the system just after an arbtrary slot boundary z B z P * ( )( ) ( )) ( ρ ) z Generatng Fn. of the number of slots spent by a job watng n the queue, pror to servce G Wq ( ρ)( [ )] ( λ [ ) z][ ] Copyrght 2002, Sanjay K. Bose 20 0

Mean number n system just after an arbtrary slot boundary (or at any arbtrary tme nstant) 2 λ b N λρ + λ 2( ρ) b + ρ Mean queueng delay (number λ b λρ + λ W of slots) encountered by a job q 2( ρ) 2 b Copyrght 2002, Sanjay K. Bose 2 The Geo [X} /G/ Queue (Early Arrval Model) Generatng Fn. of the number n the system just after a servce completon Generatng Fn. of the number n the system just after an arbtrary slot boundary ( ρ)[ ] ) λ [ ) z] ( )( ) P * ( ρ ) z Generatng Fn. of the number of slots spent by a job watng n the queue, pror to servce G Wq ( ρ)( [ )] ( λ [ ) z][ ] Note that P * ( and G W q ( are the same as for the Late Arrval Model The queue performance parameters N, N q, W, W q wll be the same as for the late arrval model Copyrght 2002, Sanjay K. Bose 22