Priority Queuing with Finite Buffer Size and Randomized Push-out Mechanism

Similar documents
Queueing Networks II Network Performance

Hidden Markov Model Cheat Sheet

Algorithms for factoring

Birth Death Processes

Fuzzy approach to solve multi-objective capacitated transportation problem

Assignment 5. Simulation for Logistics. Monti, N.E. Yunita, T.

Average Decision Threshold of CA CFAR and excision CFAR Detectors in the Presence of Strong Pulse Jamming 1

Traffic Management for the Gentle Random Early Detection using Discrete- Time Queueing

ECE697AA Lecture 17. Birth-Death Processes

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

FEATURE ANALYSIS ON QUEUE LENGTH OF ASYMMETRIC TWO-QUEUE POLLING SYSTEM WITH GATED SERVICES *

Confidence intervals for weighted polynomial calibrations

CIS526: Machine Learning Lecture 3 (Sept 16, 2003) Linear Regression. Preparation help: Xiaoying Huang. x 1 θ 1 output... θ M x M

TCP NewReno Throughput in the Presence of Correlated Losses: The Slow-but-Steady Variant

Analysis of Discrete Time Queues (Section 4.6)

An Upper Bound on SINR Threshold for Call Admission Control in Multiple-Class CDMA Systems with Imperfect Power-Control

Module 3 LOSSY IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur

Network of Markovian Queues. Lecture

Departure Process from a M/M/m/ Queue

CHARACTERISTICS OF COMPLEX SEPARATION SCHEMES AND AN ERROR OF SEPARATION PRODUCTS OUTPUT DETERMINATION

Approccio Statistico all'analisi di Sistemi Caotici e Applicazioni all'ingegneria dell'informazione

2-Adic Complexity of a Sequence Obtained from a Periodic Binary Sequence by Either Inserting or Deleting k Symbols within One Period

An application of generalized Tsalli s-havrda-charvat entropy in coding theory through a generalization of Kraft inequality

Timing-Driven Placement. Outline

Michael Batty. Alan Wilson Plenary Session Entropy, Complexity, & Information in Spatial Analysis

Hybrid Unicast and Multicast Flow Control: A Linear Optimization Approach

Tokyo Institute of Technology Periodic Sequencing Control over Multi Communication Channels with Packet Losses

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

Classification as a Regression Problem

COMPARISON OF SOME RELIABILITY CHARACTERISTICS BETWEEN REDUNDANT SYSTEMS REQUIRING SUPPORTING UNITS FOR THEIR OPERATIONS

Lesson 16: Basic Control Modes

Pricing Network Services by Jun Shu, Pravin Varaiya

A Lower Bound on SINR Threshold for Call Admission Control in Multiple-Class CDMA Systems with Imperfect Power-Control

Industrial Control and Monitoring

FUZZY FINITE ELEMENT METHOD

Lecture 4: November 17, Part 1 Single Buffer Management

A Mathematical Theory of Communication. Claude Shannon s paper presented by Kate Jenkins 2/19/00

Queuing system theory

The Tangential Force Distribution on Inner Cylinder of Power Law Fluid Flowing in Eccentric Annuli with the Inner Cylinder Reciprocating Axially

Markov model for analysis and modeling of Distributed Coordination Function of Multirate IEEE Mateusz Wielgosz

Bayesian Decision Theory

The Two-scale Finite Element Errors Analysis for One Class of Thermoelastic Problem in Periodic Composites

Externalities in wireless communication: A public goods solution approach to power allocation. by Shrutivandana Sharma

Parametric fractional imputation for missing data analysis. Jae Kwang Kim Survey Working Group Seminar March 29, 2010

Linear Approximation with Regularization and Moving Least Squares

6. Hamilton s Equations

ECE559VV Project Report

829. An adaptive method for inertia force identification in cantilever under moving mass

A Simple Inventory System

Distributed Optimal TXOP Control for Throughput Requirements in IEEE e Wireless LAN

Performance Evaluation of Deadline Monotonic Policy over protocol

Topology optimization of plate structures subject to initial excitations for minimum dynamic performance index

Parameter Estimation for Dynamic System using Unscented Kalman filter

PARTICIPATION FACTOR IN MODAL ANALYSIS OF POWER SYSTEMS STABILITY

Joint Energy Management and Resource Allocation in Rechargable Sensor Networks

Tensor Smooth Length for SPH Modelling of High Speed Impact

Combinational Circuit Design

An identification algorithm of model kinetic parameters of the interfacial layer growth in fiber composites

Source-Channel-Sink Some questions

SUPPLEMENTARY INFORMATION

Modelling BLUE Active Queue Management using Discrete-time Queue

Note 10. Modeling and Simulation of Dynamic Systems

Thermal-Fluids I. Chapter 18 Transient heat conduction. Dr. Primal Fernando Ph: (850)

Topic 5: Non-Linear Regression

Global Sensitivity. Tuesday 20 th February, 2018

Pricing and Resource Allocation Game Theoretic Models

THE EFFECT OF TORSIONAL RIGIDITY BETWEEN ELEMENTS ON FREE VIBRATIONS OF A TELESCOPIC HYDRAULIC CYLINDER SUBJECTED TO EULER S LOAD

EVALUATION OF THE VISCO-ELASTIC PROPERTIES IN ASPHALT RUBBER AND CONVENTIONAL MIXES

ENV05: METROLOGY FOR OCEAN SALINITY AND ACIDITY

Chapter 7 Channel Capacity and Coding

Power law and dimension of the maximum value for belief distribution with the max Deng entropy

CIS 700: algorithms for Big Data

Application of artificial intelligence in earthquake forecasting

Consistency & Convergence

Comparing two Quantiles: the Burr Type X and Weibull Cases

Lecture Notes on Linear Regression

Fast Simulation of Pyroshock Responses of a Conical Structure Using Rotation-Superposition Method

COMBINED CONTROL OF HYDROPNEUMATIC SPRINGS OF THE VEHICLE

CHAPTER 5 NUMERICAL EVALUATION OF DYNAMIC RESPONSE

Reliability Gain of Network Coding in Lossy Wireless Networks

TLCOM 612 Advanced Telecommunications Engineering II

Meenu Gupta, Man Singh & Deepak Gupta

Convexity preserving interpolation by splines of arbitrary degree

Amplification and Relaxation of Electron Spin Polarization in Semiconductor Devices

VQ widely used in coding speech, image, and video

Independent Component Analysis

Flexible Allocation of Capacity in Multi-Cell CDMA Networks

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

Entropy of Markov Information Sources and Capacity of Discrete Input Constrained Channels (from Immink, Coding Techniques for Digital Recorders)

Continuous Time Markov Chain

9.2 Seismic Loads Using ASCE Standard 7-93

Statistical Energy Analysis for High Frequency Acoustic Analysis with LS-DYNA

Lecture # 02: Pressure measurements and Measurement Uncertainties

Research Article The Point Zoro Symmetric Single-Step Procedure for Simultaneous Estimation of Polynomial Zeros

Adaptive Dynamical Polling in Wireless Networks

Inexact Newton Methods for Inverse Eigenvalue Problems

Solution of Equilibrium Equation in Dynamic Analysis. Mode Superposition. Dominik Hauswirth Method of Finite Elements II Page 1

Numerical studies of space filling designs: optimization algorithms and subprojection properties

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

Transfer Functions. Convenient representation of a linear, dynamic model. A transfer function (TF) relates one input and one output: ( ) system

Transcription:

ICN 00 Prorty Queung wth Fnte Buffer Sze and Randomzed Push-out Mechansm Vladmr Zaborovsy, Oleg Zayats, Vladmr Muluha Polytechncal Unversty, Sant-Petersburg, Russa Arl 4, 00

Content I. Introducton II. III. IV. Comuter Networ as the Prorty Queung: Throughut vs. Delay Factors Mathematcal And Structural Model of the Comuter Networ as the Prorty Queung Results of Model-based Analyss and Future Develoment V. Concluson

I. Performance Measure Characterstcs at Dfferent Layers From the very begnnng the usage of the mechansm of statstcal multlexng has rovded success to the Internet roect. Router load, Pacet loss rate, etc. Performance means: bts er second destnaton τ Traffc Fnte buffer source Prorty acets Bacground traffc Internetworng characterstcs Performance means: Throughut, Delay ( τ) Quantty of the connected users, Communcaton channels load level Performance means: number of users, alcatons and etc. 3

II. Networ Performance : Influence Factors Networ structure Networ envronment factors that reduce traffc erformance: Informaton system Alcaton secfc control congeston rocessng and queung delays Logcal structure TCP connectons Payload ACK Dynamc system n stochastc envronment Inut λ networ connectons outut IP acets Stochastc system and envronment N acets Physcal sgnal Physcal bnary sgnals Electromagnetc nose level level 0 acet loss buffer overflow Vrtual channel erformance factors :. Number of delvered acets er sec. Source-destnaton acets delvery delay 4

II. Vrtual Channel Performance: Throughut and Pacet Delay Asects Source Destnaton Pacets n vrtual channel Increase throughut - ncrease number of acets n vrtual transort channel N Pacets n heavy loaded channel Lttle s law N = λ T. By ncreasng average number of acets (N) and eeng arrval rate (λ), bandwdth and erformance of acet swtchers - we ncrease average tme acet sends n vrtual channel ( T), where t s erformance measure tme, t/ T>> Challenge : Tradeoff between Throughut (N/ t) and acet delay factor ( T). 5

III. The Sace Exerment "Contour Aboard ISS λ λ µ ρ = 0, ρ =, 6

III. Smle Model of Telematcs Devce n Comuter Networ Source Source Vdeo stream Comuter Networ Destnaton Destnaton λ λ acet fnte buffer µ Control acets α Pacet loss α For ρ =λ /µ=0. and ρ =λ /µ=. : P loss <0,0, P loss <0, : 0,695<α<0,787 - reemtve rorty 0,884<α<0,886 non-reemtve rorty 7

III. Formal Model of Telematcs Devce Extended Kendal s classfcaton: M M // / / f M M Exonental dstrbuted arrval tme of two customers tye Exonental servce tme dstrbuton α One server Fnte buffer sze f ab a rorty tye ( - reemtve rorty) b ush-out mechansm ( - randomzed) 8

III. Proosed Model Use Condtons P loss λ λ >,5 µ ~lnear λ λ < µ ~lnear 0 M / M // / f M M / M?// / f / M // f α=0 analytcal soluton 0<α< µ < λ λ <,5 µ α / α= analytcal soluton 9

III. System M /M///f State Grah Number of non-rorty acets n buffer Number of rorty acets n buffer. The number of rortes determnes the matrx dmenson (rorty tyes number). The buffer sze s the matrx sze 3. The randomzed ush-out mechansm determnes the robablty of a relacement mechansm α Numercal soluton comlexty ~ rorty tyes number Trangular matrx x wth ()/ entres Prevent from gettng analytcal soluton 0

III. Analytcal Calculatons ), (, ) ( ] ) [( ) (,, 0 = = = ρ α ξ ξ ρ α ζ ρ ρ ), 0, ; 0, ( 0, )] ( ) ( ) ( ) ( [,,,,,,0,,0,0,,,, = = = αλ δ λ λ µδ µ δ δ µ δ λ δ δ αλ δ λ System of ()/ lnear equatons System of lnear equatons Usng the method of generatng functons and some arorstc data about our system we can receve the dstrbuton of the generatng functon and calculate the characterstcs of the ntal system.

IV. Comarson of the P loss Grahs Wth Nonreemtve Prorty Queung ρ = 0, ρ =, Grahs are close to lnear curves α stronger nfluences the grahs, than rorty tye ρ =, ρ = 0, Grahs aren t lnear curves The rorty tye doesn t nfluence the grahs P loss vary from 0,03 to,83*0- P loss ncreased for % from 0, to 0,

IV. Results : Relatve Throughut of the Traffc ӕ ab = ӕ a / ӕ b = ( P (a) loss ) / ( P(b) loss ) Low loaded networ (ρ =0,5) Heavy loaded networ (ρ =,5) Monotonous curves In low loaded networ curves are close to lnear ρ =0, ρ =0,5 3 ρ =,0 4 ρ =,3 5 ρ =,5 6 ρ =,7 7 ρ =,9 8 ρ =,5 Curves wth extremums There s otmum value of a when non-rorty traffc has the lowest loss rate 3

IV. Results : The Tme That Pacets Send n Queung ( ) s n queue θ = = ρ, =, ( =,). ( ) τ τ ( Ploss ) λ Prorty acets Non-rorty acets The relatve tme that acets send n queung does not exceed some value For low loaded networs t s close to zero ρ =0, ρ =0,5 3 ρ =,0 4 ρ =,3 5 ρ =,5 6 ρ =,7 7 ρ =,9 8 ρ =,5 The relatve tme that acets send n queung fast grows wth loadng growth In heavy loaded networs t can be used to bloc undesrable or suscous traffc 4

IV. Future Develoment. Telematcs devce 3D sace 4. Practcal ntroducton. Influence of the sze of the buffer on the system s characterstcs and traffc locng 3. Herarchcal rorty 5

V. Concluson Offered comutng algorthm allows networ engneers and desgners to estmate ossble varants of networ traffc load. In overload networs, then λ λ >µ, the deendence of loss rorty from varable s not lnear and even not monotonous functon. One of the real ractcal alcaton for our model s sace exerment Contour. Proosed ossble ways to mrove model are: Varable buffer sze ; More comlex, herarchcal rorty. 6

Than you for your attenton Questons? 7