PAPER 39 STOCHASTIC NETWORKS

Similar documents
Control Chart Analysis of E k /M/1 Queueing Model

Temporal-Difference Learning

EN40: Dynamics and Vibrations. Midterm Examination Thursday March

Suggested Solutions to Homework #4 Econ 511b (Part I), Spring 2004

Efficiency Loss in a Network Resource Allocation Game: The Case of Elastic Supply

16 Modeling a Language by a Markov Process

4/18/2005. Statistical Learning Theory

MASSACHUSETTS INSTITUTE OF TECHNOLOGY Department of Electrical Engineering and Computer Science

1D2G - Numerical solution of the neutron diffusion equation

On the Control of Fork-Join Networks

Tradable Network Permits: A New Scheme for the Most Efficient Use of Network Capacity

EXAM NMR (8N090) November , am

Analysis of Loss Networks with Routing

A hybrid cross entropy algorithm for solving dynamic transit network design problem

Bayesian Congestion Control over a Markovian Network Bandwidth Process

Markscheme May 2017 Calculus Higher level Paper 3

The evolution of the phase space density of particle beams in external fields

Efficiency Loss in a Network Resource Allocation Game

CSCE 478/878 Lecture 4: Experimental Design and Analysis. Stephen Scott. 3 Building a tree on the training set Introduction. Outline.

ON INDEPENDENT SETS IN PURELY ATOMIC PROBABILITY SPACES WITH GEOMETRIC DISTRIBUTION. 1. Introduction. 1 r r. r k for every set E A, E \ {0},

EM Boundary Value Problems

Multipath routing and congestion control

Conventional Paper-I (a) Explain the concept of gradient. Determine the gradient of the given field: ( )

3.1 Random variables

Goodness-of-fit for composite hypotheses.

Absolute Specifications: A typical absolute specification of a lowpass filter is shown in figure 1 where:

New problems in universal algebraic geometry illustrated by boolean equations

The Substring Search Problem

11) A thin, uniform rod of mass M is supported by two vertical strings, as shown below.

International Journal of Mathematical Archive-3(12), 2012, Available online through ISSN

Research Article On Alzer and Qiu s Conjecture for Complete Elliptic Integral and Inverse Hyperbolic Tangent Function

MULTILAYER PERCEPTRONS

Internet Appendix for A Bayesian Approach to Real Options: The Case of Distinguishing Between Temporary and Permanent Shocks

Modelling packet insertion on a WSADM ring

Modelling packet insertion on a WSADM ring

An Application of Fuzzy Linear System of Equations in Economic Sciences

Stochastic Analysis of Periodic Real-Time Systems *

arxiv: v2 [math.pr] 14 Jun 2016

6 PROBABILITY GENERATING FUNCTIONS

Multiple Criteria Secretary Problem: A New Approach

School of Electrical and Computer Engineering, Cornell University. ECE 303: Electromagnetic Fields and Waves. Fall 2007

Parameter identification in Markov chain choice models

763620SS STATISTICAL PHYSICS Solutions 2 Autumn 2012

A Multivariate Normal Law for Turing s Formulae

Bifurcation Analysis for the Delay Logistic Equation with Two Delays

On Computing Optimal (Q, r) Replenishment Policies under Quantity Discounts

Absorption Rate into a Small Sphere for a Diffusing Particle Confined in a Large Sphere

Homework 7 Solutions

DonnishJournals

A DETRMINISTIC RELIABILITY BASED MODEL FOR PROCESS CONTROL

Is there a magnification paradox in gravitational lensing?

Revisiting Stochastic Loss Networks: Structures and Algorithms

Magnetic Field. Conference 6. Physics 102 General Physics II

Regularization. Stephen Scott and Vinod Variyam. Introduction. Outline. Machine. Learning. Problems. Measuring. Performance.

Current Balance Warm Up

COLLISIONLESS PLASMA PHYSICS TAKE-HOME EXAM

Numerical Integration

SAMPLE QUIZ 3 - PHYSICS For a right triangle: sin θ = a c, cos θ = b c, tan θ = a b,

MATH 415, WEEK 3: Parameter-Dependence and Bifurcations

The second law of thermodynamics - II.

6 Matrix Concentration Bounds

AST 121S: The origin and evolution of the Universe. Introduction to Mathematical Handout 1

Rate Splitting is Approximately Optimal for Fading Gaussian Interference Channels

Validity of heavy-traffic steady-state approximations in multiclass queueing networks: The case of queue-ratio disciplines

Interaction of Feedforward and Feedback Streams in Visual Cortex in a Firing-Rate Model of Columnar Computations. ( r)

Macro Theory B. The Permanent Income Hypothesis

Fair Internet traffic integration: network flow models and analysis

DEMONSTRATION OF INADEQUACY OF FFOWCS WILLIAMS AND HAWKINGS EQUATION OF AEROACOUSTICS BY THOUGHT EXPERIMENTS. Alex Zinoviev 1

Control of parallel non-observable queues: asymptotic equivalence and optimality of periodic policies

Prediction of Motion Trajectories Based on Markov Chains

arxiv: v1 [quant-ph] 15 Nov 2018

DOING PHYSICS WITH MATLAB COMPUTATIONAL OPTICS

10.2 Parametric Calculus

A Bijective Approach to the Permutational Power of a Priority Queue

ECE 3318 Applied Electricity and Magnetism. Spring Prof. David R. Jackson ECE Dept. Notes 13

An extended target tracking method with random finite set observations

Anyone who can contemplate quantum mechanics without getting dizzy hasn t understood it. --Niels Bohr. Lecture 17, p 1

Pushdown Automata (PDAs)

Control of Patient Flow in Emergency Departments, or Multiclass Queues with Deadlines and Feedback

CENTER FOR MULTIMODAL SOLUTIONS FOR CONGESTION MITIGATION (CMS)

Math 124B February 02, 2012

Queuing Network Approximation Technique for Evaluating Performance of Computer Systems with Hybrid Input Source

, the tangent line is an approximation of the curve (and easier to deal with than the curve).

A NEW VARIABLE STIFFNESS SPRING USING A PRESTRESSED MECHANISM

Bounds on the performance of back-to-front airplane boarding policies

University of Illinois at Chicago Department of Physics. Electricity & Magnetism Qualifying Examination

Magnetic Dipoles Challenge Problem Solutions

Aalborg Universitet. Load Estimation from Natural input Modal Analysis Aenlle, Manuel López; Brincker, Rune; Canteli, Alfonso Fernández

MU+CU+KU=F MU+CU+KU=0

Random Variables and Probability Distribution Random Variable

0606 ADDITIONAL MATHEMATICS 0606/01 Paper 1, maximum raw mark 80

PROBLEM SET #1 SOLUTIONS by Robert A. DiStasio Jr.

Adaptive Checkpointing in Dynamic Grids for Uncertain Job Durations

A Comment on Increasing Returns and Spatial. Unemployment Disparities

A New Approach to General Relativity

Handout: IS/LM Model

Velocimetry Techniques and Instrumentation

Localization of Eigenvalues in Small Specified Regions of Complex Plane by State Feedback Matrix

Pervasive and Mobile Computing. Analysis of power saving and its impact on web traffic in cellular networks with continuous connectivity

A Power Method for Computing Square Roots of Complex Matrices

Transcription:

MATHEMATICAL TRIPOS Pat III Tuesday, 2 June, 2015 1:30 pm to 4:30 pm PAPER 39 STOCHASTIC NETWORKS Attempt no moe than FOUR questions. Thee ae FIVE questions in total. The questions cay equal weight. STATIONERY REQUIREMENTS Cove sheet Teasuy Tag Scipt pape SPECIAL REQUIREMENTS None You may not stat to ead the questions pinted on the subsequent pages until instucted to do so by the Invigilato.

1 2 Pove that in equilibium the depatue pocess fom an M/M/K queue is a Poisson pocess. Ailine passenges aive at passpot contol in accodance with a Poisson pocess of ate ν. The contol opeates as a two seve queue at which sevice times ae independent and exponentially distibuted with mean µ < 2/ν and ae independent of the aival pocess. Afte leaving passpot contol, a passenge must pass though a secuity check. This opeates as a single-seve queue at which sevice times ae all equal to a constant τ(< ν 1 ). Show that in equilibium the pobability both queues ae empty is 2 νµ 2+νµ (1 ντ). Ifit takes atimeσ towalk fomthefistqueuetothesecond, whatis theequilibium pobability that both queues ae empty and thee is no passenge walking between them? 2 Deive Elang s fomula, E(ν,C), fo the popotion of calls lost at a esouce of capacity C offeed a load of ν. State clealy any assumptions you make in you deivation. Define a loss netwok with a fixed outing, and descibe biefly how the Elang fixed point equations ( ) B j = E (1 B j ) 1 A j ν (1 B i ) A i,c j, j = 1,2,...J (1) i aise as a natual appoximation fo the link blocking pobabilities in the netwok, whee A is the link-oute incidence matix, assumed to be a 0 1 matix. Show that a solution to the equations (1) also locates the minimum of a stictly convex function ν e j y ja j + yj U(z,C j )dz j 0 ove the positive othant y 0, fo a function U to be detemined. Descibe a fom of epeated substitution that conveges to a solution of equations (1), and pove that it conveges.

3 3 Define a Wadop equilibium fo the flows x = (x, R) in a congested netwok with outes R and links j J. Show that if the delay D j (y j ) at link j is a continuously diffeentiable, stictly inceasing function of the thoughput, y j, of the link j then a Wadop equilibium exists and solves an optimization poblem of the fom minimize ove j J yj 0 x 0, y D j (u)du subject to Hx = f, Ax = y, whee f = (f s,s S) and f s is the (fixed) aggegate flow between souce-sink pai s. What ae the matices A and H? Ae the equilibium thoughputs, y j, unique? Ae the equilibium flows, x, unique? Justify you answes. Suppose now that, in addition to the delay D j (y j ), uses of link j incu a tafficdependent toll T j (y j ), and suppose each use selects a oute in an attempt to minimize the sum of its toll and its delays. Show that it is possible to choose the functions T j ( ) so that the equilibium flow patten minimizes the aveage delay in the netwok. [TURN OVER

4 4 The dynamical system ( ) d dt µ j(t) = κ j µ j (t) A j x (t) C j j J x (t) = w k µ k(t)a k R is poposed as a model fo esouce allocation in a netwok, whee R is a set of outes, J is a set of esouces, A is a 0 1 matix, w > 0 fo R, C j is the capacity of esouce j and κ j > 0 fo j J. Povide a bief intepetation of this model as a pocess that attempts to balance supply and demand. By consideing the function V(µ) = Rw log j µ j A j µ j C j j J o othewise, show that if the matix A has full ank then all tajectoies of the dynamical system convege towad a uniqueequilibium point. What happensif A is not of full ank?

5 5 Let J be a set of esouces, and R a set of outes, whee a oute R identifies a subset of J. Let C j be the capacity of esouce j, and suppose the numbe of flows in pogess on each oute is given by the vecto n = (n, R). Define a popotionally fai ate allocation. Conside a netwok with esouces J = {1, 2}, each of unit capacity, and outes R = {{1},{2},{1,2}}. Given n = (n, R), find the ate x of each flow on oute, fo each R, unde a popotionally fai ate allocation. Suppose now that flows descibe the tansfe of documents though a netwok, that new flows oiginate as independent Poisson pocesses of ates ν, R, and that document sizes ae independent and exponentially distibuted with paamete µ fo each oute R. Detemine the tansition intensities of the esulting Makov pocess n = (n, R). Show that the stationay distibution of the Makov pocess n = (n, R) takes the fom ( ) n{1} +n {2} +n {1,2} ( ) n ν π(n) = B, n {1,2} whee B is a nomalizing constant, povided the paametes (ν,µ, R) satisfy cetain conditions. Detemine these conditions, and also the constant B. Show that, unde the distibution π, n {1} and n {2} ae independent. Ae n {1}, n {2} and n {1,2} independent? R µ END OF PAPER