CS/ECE 715 Spring 2004 Homework 5 (Due date: March 16)

Similar documents
Queuing Theory. Basic properties, Markovian models, Networks of queues, General service time distributions, Finite source models, Multiserver queues

First come, first served (FCFS) Batch

Reliability and Queueing

A queueing system can be described as customers arriving for service, waiting for service if it is not immediate, and if having waited for service,

Queueing theory and Replacement model

B. Maddah ENMG 622 ENMG /27/07


Statistical Inference (Chapter 10) Statistical inference = learn about a population based on the information provided by a sample.

Generalized Semi- Markov Processes (GSMP)

1 Introduction to reducing variance in Monte Carlo simulations

Simulation. Two Rule For Inverting A Distribution Function

Announcements. Queueing Systems: Lecture 1. Lecture Outline. Topics in Queueing Theory

c. Explain the basic Newsvendor model. Why is it useful for SC models? e. What additional research do you believe will be helpful in this area?

SINGLE-CHANNEL QUEUING PROBLEMS APPROACH

The picture in figure 1.1 helps us to see that the area represents the distance traveled. Figure 1: Area represents distance travelled

Discrete probability distributions

Limit Theorems. Convergence in Probability. Let X be the number of heads observed in n tosses. Then, E[X] = np and Var[X] = np(1-p).

STAT 516 Answers Homework 6 April 2, 2008 Solutions by Mark Daniel Ward PROBLEMS

Simulation of Discrete Event Systems

Birth-Death Processes. Outline. EEC 686/785 Modeling & Performance Evaluation of Computer Systems. Relationship Among Stochastic Processes.

This section is optional.

TCOM 501: Networking Theory & Fundamentals. Lecture 3 January 29, 2003 Prof. Yannis A. Korilis

Output Analysis (2, Chapters 10 &11 Law)

Randomized Algorithms I, Spring 2018, Department of Computer Science, University of Helsinki Homework 1: Solutions (Discussed January 25, 2018)

Control chart for number of customers in the system of M [X] / M / 1 Queueing system

Discrete Mathematics for CS Spring 2008 David Wagner Note 22

ECE 901 Lecture 12: Complexity Regularization and the Squared Loss

There is no straightforward approach for choosing the warmup period l.

HOMEWORK I: PREREQUISITES FROM MATH 727

Massachusetts Institute of Technology

TUFTS UNIVERSITY DEPARTMENT OF CIVIL AND ENVIRONMENTAL ENGINEERING ES 152 ENGINEERING SYSTEMS Spring Lesson 14. Queue System Theory

Problems from 9th edition of Probability and Statistical Inference by Hogg, Tanis and Zimmerman:

Probability and statistics: basic terms

EE 4TM4: Digital Communications II Probability Theory

NOTES ON DISTRIBUTIONS

Estimation for Complete Data

Chapter 11 Output Analysis for a Single Model. Banks, Carson, Nelson & Nicol Discrete-Event System Simulation

Expectation and Variance of a random variable

Final Review for MATH 3510

Topic 9: Sampling Distributions of Estimators

Queueing Theory II. Summary. M/M/1 Output process Networks of Queue Method of Stages. General Distributions

Stat 400: Georgios Fellouris Homework 5 Due: Friday 24 th, 2017

1 Convergence in Probability and the Weak Law of Large Numbers

MATH 320: Probability and Statistics 9. Estimation and Testing of Parameters. Readings: Pruim, Chapter 4

6.041/6.431 Spring 2009 Final Exam Thursday, May 21, 1:30-4:30 PM.

STAT 350 Handout 19 Sampling Distribution, Central Limit Theorem (6.6)

Discrete Mathematics and Probability Theory Spring 2016 Rao and Walrand Note 19

4. Basic probability theory

Introduction to probability Stochastic Process Queuing systems. TELE4642: Week2

Chapter 6 Principles of Data Reduction

Quick Review of Probability

Lecture 6: Coupon Collector s problem

Exercise 4.3 Use the Continuity Theorem to prove the Cramér-Wold Theorem, Theorem. (1) φ a X(1).

Approximations and more PMFs and PDFs

Quick Review of Probability

EECS564 Estimation, Filtering, and Detection Hwk 2 Solns. Winter p θ (z) = (2θz + 1 θ), 0 z 1

ECE 6980 An Algorithmic and Information-Theoretic Toolbox for Massive Data

The variance of a sum of independent variables is the sum of their variances, since covariances are zero. Therefore. V (xi )= n n 2 σ2 = σ2.

AMS570 Lecture Notes #2

Problem Set 4 Due Oct, 12

Math 155 (Lecture 3)

HOMEWORK #10 SOLUTIONS

Since X n /n P p, we know that X n (n. Xn (n X n ) Using the asymptotic result above to obtain an approximation for fixed n, we obtain

THE ROYAL STATISTICAL SOCIETY GRADUATE DIPLOMA STATISTICAL THEORY AND METHODS PAPER I

Confidence Level We want to estimate the true mean of a random variable X economically and with confidence.

First Year Quantitative Comp Exam Spring, Part I - 203A. f X (x) = 0 otherwise

This exam contains 19 pages (including this cover page) and 10 questions. A Formulae sheet is provided with the exam.

Monte Carlo Integration

7.1 Convergence of sequences of random variables

Direction: This test is worth 150 points. You are required to complete this test within 55 minutes.

Lecture 7: Properties of Random Samples

The statistical pattern of the arrival can be indicated through the probability distribution of the number of the arrivals in an interval.

Topic 5 [434 marks] (i) Find the range of values of n for which. (ii) Write down the value of x dx in terms of n, when it does exist.

Application to Random Graphs

Topic 9: Sampling Distributions of Estimators

0, otherwise. EX = E(X 1 + X n ) = EX j = np and. Var(X j ) = np(1 p). Var(X) = Var(X X n ) =

Increasing timing capacity using packet coloring

MATH 2300 review problems for Exam 2

Table 12.1: Contingency table. Feature b. 1 N 11 N 12 N 1b 2 N 21 N 22 N 2b. ... a N a1 N a2 N ab

Homework 3 Solutions

Summary. Recap ... Last Lecture. Summary. Theorem

CSE 4095/5095 Topics in Big Data Analytics Spring 2017; Homework 1 Solutions

Discrete Mathematics for CS Spring 2007 Luca Trevisan Lecture 22

The standard deviation of the mean

STAT Homework 1 - Solutions

A sequence of numbers is a function whose domain is the positive integers. We can see that the sequence

Stat410 Probability and Statistics II (F16)

Stat 319 Theory of Statistics (2) Exercises

Chapter 20. Comparing Two Proportions. BPS - 5th Ed. Chapter 20 1

Solutions to Homework 6

Resampling Methods. X (1/2), i.e., Pr (X i m) = 1/2. We order the data: X (1) X (2) X (n). Define the sample median: ( n.

CEE 522 Autumn Uncertainty Concepts for Geotechnical Engineering

B. Maddah ENMG 622 ENMG /20/09

Kinetics of Complex Reactions

Chapter 2 The Monte Carlo Method

Discrete Mathematics and Probability Theory Summer 2014 James Cook Note 15

Castiel, Supernatural, Season 6, Episode 18

Closed book and notes. No calculators. 60 minutes, but essentially unlimited time.

Homework 5 Solutions

Convergence of random variables. (telegram style notes) P.J.C. Spreij

Transcription:

CS/ECE 75 Sprig 004 Homework 5 (Due date: March 6) Problem 0 (For fu). M/G/ Queue with Radom-Sized Batch Arrivals. Cosider the M/G/ system with the differece that customers are arrivig i batches accordig to a Poisso process with rate λ. Each batch has customers, where has a give distributio ad is idepedet of customer service times. Show that the waitig time i queue is give by λx X W ------------------- () ( ρ) + ----------------------- ( ρ) where X deotes service time. [Solutio] The waitig time of a customer, W, cosists of three parts, the residual service time of the customer beig served R, the average service time of all customers waitig i queue ρw, ad the average waitig time of the customer for other customers who arrived i the same batch. where the mea residual service time is W R + ρw + W B () λx R ------------ (3) We derive the average waitig time of a customer for other customers that arrived i the same batch where W B j r j E{ W B batch has size j} (4) P j Probability a batch has size j (5) r j Proportio of customers arrivig i a batch of size j (6) We have Also sice the customer is equally likely to be i ay positio withi the batch Thus r j jp -------------------- j P jp ------ j j E{ W B batch is of size j} ( k )X -- j --------- X j k (7) (8)

E{ W B } Substitutig i () we obtai (). Problem. I the M/G/ system show that jp j ( j )X -------------------------- j X ( ) ---------------------- (9) P( the system is empty) λx Average legth of time betwee busy periods -- λ X Average legth of busy period --------------- λx Average umber of customers served i a busy period --------------- λx where X is average service time. [Solutio] From Little's Theorem we have that P{ busy} λx. Therefore P{ idle} λx. The legth of a idle period is the remaiig iterarrival time. Therefore it has a expoetial distributio with parameter λ. Let B be the average legth of a busy period ad let I be the average legth of a idle period. By expressig the proportio of time the system is busy as B/(I+B) ad also as B X ( λx). From this the expressio period is evidet. ( λx) we obtai for the average umber of customers served i a busy λx Problem. Cosider a computer system with a CPU ad oe disk drive. After a burst at the CPU the job completes executio with probability 0. ad request a disk I/O with probability 0.9. The time of a sigle CPU burst is expoetially distributed with mea 0.0 s. The disk service time is broke up ito three phases: expoetially distributed seek time with mea 0.03 s, uiformly distributed latecy time with mea 0.0 s, ad a costat trasfer time equal to 0.0 s. After a service completio at the disk, the job always requires a CPU burst. The average arrival rate of jobs is 0.8 job/s ad the system does ot have eough mai memory to support multiprogrammig. Solve for the average respose time. I order to computer the mea ad the variace of the service time distributio, you may eed the followig results o radom sums. [Radom Sums] We have cosidered sums of N mutually idepedet radom variable whe N is a fixed costat. Here we are iterested i the case whe N itself is radom variable that is idepedet of

X k. Give a list X, X,, of mutually i.i.d. radom variables with distributio fuctio Fx ( ), mea EX [ ] ad variace Var[ X], cosider the radom sum T X + X + + X N. Here the pmf of the discrete radom variable p N ( ) is assumed to be give. For a fixed value N, the coditioal expectatio of T is easily obtaied ETN [ ] EX [ i ] The, usig the theorem of total expectatio, we get i E[ X] (0) Similarly, we have ET [ ] E[ X]p N ( ) EX [ ] p N ( ) EX [ ]EN [ ] () Var[ T] ET [ ] ( ET [ ]) E{ E[ T N ] } ( ET [ ]) () E{ Var[ X] + ( EX [ ]) } ( ET [ ]) Var[ X]EN [ ] + ( EX [ ]) Var[ N] [Solutio] From the give parameters, we have: λ 0.8, mea time of a CPU burst ET [ c ] 0.0, mea disk seek time ET [ s ] 0.03, mea disk latecy time ET [ l ] 0.0, ad mea trasfer time ET [ t ] 0.0. Besides, the job requests a disk with probability p 0.9. Sice ad are expoetially distributed, T c T s ET [ s ] ( ET [ s ]) 8 0 4 [ ] ( ET [ c ]) 0 4 Sice is uiformly distributed i iterval (0,0.0), T l ET c (3) 0.0 3 4 ET [ l ] ------------------- -- 0 4 (4) 3 0.0 3 Sice T t is a costat time, ET [ t ] 0 4, the, the disk service time T d T s + T l + T t, we get ET [ d ] ET [ s ] + ET [ l ] + ET [ t ] 0.05 (5) ET [ d ] E[ ( T s + T l + T t ) ] 3.433 0 3 (6) From the descriptio of the problem, the total service time T is it c + ( i )T d with probability p ( i ) ( p), i,,,. Thus ET [ ] p ( i ) ( p) [ ie[ T c ] + ( i )ET [ d ]] i ET [ c ] ------------- ip i 9 + 0. ET [ d ] ip i 0 ET [ c ] + 9 ET [ d ] 0.55 3 (7)

ET [ ] p ( i ) ( p)et [ () i ] i p ( i ) ( p) ie[ T c ] ii ( )ET [ c ]ET [ d ] ( i ) ( + + ET [ d ]) i -- ( ET [ 9 c ] ET [ c ]ET [ d ]) ip i + --E[ T 9 c ]ET [ d ] + 0. ET [ d ] i p i i i (8) time 0679 Variace: Var[ T] ET [ ] ( ET [ ]) 0.3765. So, ρ λet [ ] 0.44, the average respose ER [ ] ca be obtaied: EN [ ] λe[ T ] ER [ ] ------------ ET [ ] + -------------------.305 (9) λ ( ρ) Problem 3. Persos arrive at a Xerox machie accordig to a Poisso process with rate oe per miute. The umber of copies to be made by each perso is uiformly distributed betwee ad 0. Each copy requires 3 sec. Fid the average waitig time i queue whe: [Solutio] (a) Each perso uses the machie o a first-come first-serve basis. (b) Persos with o more tha copies to make are give preemptive priority over other persos. (a) λ 60 per secod EX ( ) 6.5 secods EX ( ) 346.5 secods T E( X) + λex ( ) [ ( λe( X) )] 0.48 secods. W T E( X) 3.98 secods (b) Preemptive Queueig λ -----, λ, 60 -------- λ 300 ----- 75 EX ( ).5, EX ( ) 47.5 λ R ----- EX ( ) 0.0375 4

R R + --λ EX ( ).8875 EX ( T )( ρ ) + R EX ( ----------------------------------------------- 4.538, T )( ρ ρ ) + R ρ ----------------------------------------------------------- ( ρ )( ρ ρ ) 3.84 W T EX ( ) 0.038, W T EX ( ) 4.34 (Assume a customer with lower priority is put back ito the queue oce the service is preempted by a customer with higher priority.) Problem 4. Priority Systems with Multiple Servers. Cosider the priority systems assumig that there are m servers ad that all priority classes have expoetially distributed service times with commo mea µ. (a) Cosider the opreemptive system. Show that the average queueig time is give by R W k ---------------------------------------------------------------------------------------------- ( ρ ρ k )( ρ ρ k ) with R, the mea residual time, is give by (0) where P Q P R ------- Q () mµ is the steady-state probability of queueig give by the Erlag C formula. [Here ρ i λ i ( mµ ) ad ρ ρ i.] i (b) Cosider the preemptive resume system. Argue that W ( k), defied as the average time i queue averaged over the first k priority classes, is the same as for a M/M/m system with arrival rate λ + λ + + λ k ad mea service time µ. Use Little s Theorem to show that the average time i queue of a k th priority class customer ca be obtaied recursively from W W ( ) k k W k ---- W, (3) λ ( k) λ i k W ( k ) λ i k 3,,, i i [Erlag C formula] () 5

I a M/M/m system, the probability that a arrival will fid all servers busy is [Solutio] (a) The same derivatio as i the otes applies for, i.e., W k R ( ρ ρ k )( ρ ρ k ) (6) where ρ i λ i ( mµ ), ad R is the mea residual service time. Thik of the system as beig comprised of a servig sectio ad a waitig sectio. The residual service time is just the time util ay of the customers i the waitig sectio ca eter the servig sectio. Thus, the residual service time of a customer is zero if the customer eters service immediately because there is a free server at the time of arrival, ad is otherwise equal to the time betwee the customer s arrival, ad the first subsequet service completio. Usig the memoryless property of the expoetial distributio it is see that p P Q P{ Queueig} 0 ( mρ) m -----------------------, where p is (4) m! ( ρ) 0 R P Q E{ Residual service time queueig occurs} P Q ( mµ ) (7) (b) The waitig time of classes,, k is ot iflueced by the presece of classes (k+),...,. All priority classes have the same service time distributio, thus, iterchagig the order of service does ot chage the average waitig time,. We have p 0 m ( mρ) -------------- ( mρ) m! + ----------------------- m! ( ρ) 0 W k (5) W ( k) Average waitig time for the M/M/m system with rate ( λ + + λ k ) By Little s theorem we have Average umber i queue of class k Average umber i queue of classes to k Average umber i queue of classes to k- ad the desired result follows. (8) (9) 6