Routing in Delay Tolerant Networks

Similar documents
UNIT #5 EXPONENTIAL AND LOGARITHMIC FUNCTIONS

CPSC 211 Data Structures & Implementations (c) Texas A&M University [ 259] B-Trees

Poisson process Markov process

Microscopic Flow Characteristics Time Headway - Distribution

Economics 302 (Sec. 001) Intermediate Macroeconomic Theory and Policy (Spring 2011) 3/28/2012. UW Madison

CSE 245: Computer Aided Circuit Simulation and Verification

Lecture 1: Numerical Integration The Trapezoidal and Simpson s Rule

Institute of Actuaries of India

AR(1) Process. The first-order autoregressive process, AR(1) is. where e t is WN(0, σ 2 )

Lecture 2: Current in RC circuit D.K.Pandey

4.1 The Uniform Distribution Def n: A c.r.v. X has a continuous uniform distribution on [a, b] when its pdf is = 1 a x b

Spring 2006 Process Dynamics, Operations, and Control Lesson 2: Mathematics Review

Ma/CS 6a Class 15: Flows and Bipartite Graphs

14.02 Principles of Macroeconomics Problem Set 5 Fall 2005

Wave Equation (2 Week)

SOLUTIONS. 1. Consider two continuous random variables X and Y with joint p.d.f. f ( x, y ) = = = 15. Stepanov Dalpiaz

Final Exam : Solutions

CHAPTER CHAPTER14. Expectations: The Basic Tools. Prepared by: Fernando Quijano and Yvonn Quijano

First Lecture of Machine Learning. Hung-yi Lee

Phys463.nb Conductivity. Another equivalent definition of the Fermi velocity is

Random Access Techniques: ALOHA (cont.)

The Science of Monetary Policy

Physics 160 Lecture 3. R. Johnson April 6, 2015

On the Derivatives of Bessel and Modified Bessel Functions with Respect to the Order and the Argument

Applied Statistics and Probability for Engineers, 6 th edition October 17, 2016

Midterm Examination (100 pts)

Chapter 12 Introduction To The Laplace Transform

Transfer function and the Laplace transformation

An Indian Journal FULL PAPER. Trade Science Inc. A stage-structured model of a single-species with density-dependent and birth pulses ABSTRACT

On the Speed of Heat Wave. Mihály Makai

The transition:transversion rate ratio vs. the T-ratio.

Network Congestion Games

A THREE COMPARTMENT MATHEMATICAL MODEL OF LIVER

Real time estimation of traffic flow and travel time Based on time series analysis

Double Slits in Space and Time

B) 25y e. 5. Find the second partial f. 6. Find the second partials (including the mixed partials) of

Power communication network traffic prediction based on two dimensional prediction algorithms Lei Xiao Xue Yang Min Zhu Lipeng Zhu

Section. Problem Representation. Substation. Protection Device. protection equipments. Substation. Client. EPDS divided in blocks connected by

Integrity Control in Nested Certificates

Mixing Real-Time and Non-Real-Time. CSCE 990: Real-Time Systems. Steve Goddard.

DEPARTMENT OF ELECTRICAL &ELECTRONICS ENGINEERING SIGNALS AND SYSTEMS. Assoc. Prof. Dr. Burak Kelleci. Spring 2018

5. An object moving along an x-coordinate axis with its scale measured in meters has a velocity of 6t

Circuits and Systems I

Midterm exam 2, April 7, 2009 (solutions)

General Article Application of differential equation in L-R and C-R circuit analysis by classical method. Abstract

Charging of capacitor through inductor and resistor

2.1. Differential Equations and Solutions #3, 4, 17, 20, 24, 35

14.02 Principles of Macroeconomics Fall 2005 Quiz 3 Solutions

Voltage v(z) ~ E(z)D. We can actually get to this wave behavior by using circuit theory, w/o going into details of the EM fields!

C From Faraday's Law, the induced voltage is, C The effect of electromagnetic induction in the coil itself is called selfinduction.

Mundell-Fleming I: Setup

Control System Engineering (EE301T) Assignment: 2

I) Title: Rational Expectations and Adaptive Learning. II) Contents: Introduction to Adaptive Learning

cs/ee 143 Communication Networks

3(8 ) (8 x x ) 3x x (8 )

= x. I (x,y ) Example: Translation. Operations depend on pixel s Coordinates. Context free. Independent of pixel values. I(x,y) Forward mapping:

3.4 Repeated Roots; Reduction of Order

Search sequence databases 3 10/25/2016

Math 34A. Final Review

1 Minimum Cut Problem

EE 434 Lecture 22. Bipolar Device Models

3+<6,&6([DP. September 29, SID (last 5 digits): --

Vehicle Arrival Models : Headway

Elementary Differential Equations and Boundary Value Problems

EEO 401 Digital Signal Processing Prof. Mark Fowler

Study on the Lightweight checkpoint based rollback recovery mechanism

More Digital Logic. t p output. Low-to-high and high-to-low transitions could have different t p. V in (t)

MEM 355 Performance Enhancement of Dynamical Systems A First Control Problem - Cruise Control

Voltage v(z) ~ E(z)D. We can actually get to this wave behavior by using circuit theory, w/o going into details of the EM fields!

Discussion 06 Solutions

Let s look again at the first order linear differential equation we are attempting to solve, in its standard form:

Economics 302 (Sec. 001) Intermediate Macroeconomic Theory and Policy (Spring 2011) 4/25/2011. UW Madison

Lecture 1: Growth and decay of current in RL circuit. Growth of current in LR Circuit. D.K.Pandey

Resource Allocation in Visible Light Communication Networks NOMA vs. OFDMA Transmission Techniques

EXERCISE - 01 CHECK YOUR GRASP

REPETITION before the exam PART 2, Transform Methods. Laplace transforms: τ dτ. L1. Derive the formulas : L2. Find the Laplace transform F(s) if.

THE SHORT-RUN AGGREGATE SUPPLY CURVE WITH A POSITIVE SLOPE. Based on EXPECTATIONS: Lecture. t t t t

Themes. Flexible exchange rates with inflation targeting. Expectations formation under flexible exchange rates

Chapter 3: Fourier Representation of Signals and LTI Systems. Chih-Wei Liu

Gradebook & Midterm & Office Hours

Decline Curves. Exponential decline (constant fractional decline) Harmonic decline, and Hyperbolic decline.

Availability Analysis of Repairable Computer Systems and Stationarity Detection

1. Inverse Matrix 4[(3 7) (02)] 1[(0 7) (3 2)] Recall that the inverse of A is equal to:

The Optimal Timing of Transition to New Environmental Technology in Economic Growth

A Propagating Wave Packet Group Velocity Dispersion

A MATHEMATICAL MODEL FOR NATURAL COOLING OF A CUP OF TEA

Modal identification of structures from roving input data by means of maximum likelihood estimation of the state space model

Chapter 17 Handout: Autocorrelation (Serial Correlation)

The Procedure Abstraction Part II: Symbol Tables and Activation Records

x(m) t(sec ) Homework #2. Ph 231 Introductory Physics, Sp-03 Page 1 of 4

where: u: input y: output x: state vector A, B, C, D are const matrices

EE 529 Remote Sensing Techniques. Review

Azimuthal angular correlations between heavy flavour decay electrons and charged hadrons in pp collisions at s = 2.76 TeV in ALICE

4.2 Design of Sections for Flexure

Reliability Analysis of a Bridge and Parallel Series Networks with Critical and Non- Critical Human Errors: A Block Diagram Approach.

ANSWERS TO EVEN NUMBERED EXERCISES IN CHAPTER 11

priority queue ADT heaps 1

H is equal to the surface current J S

Chapter 6 Folding. Folding

Boyce/DiPrima 9 th ed, Ch 2.1: Linear Equations; Method of Integrating Factors

Transcription:

Rouing in Dlay Tolran Nworks Primary Rfrnc: S. Jain K. Fall and R. Para Rouing in a Dlay Tolran Nwork SIGCOMM 04 Aug. 30-Sp. 3 2004 Porland Orgon USA Sudn lcur by: Soshan Bali (748214) mail : sbali@ic.ku.du For EECS 800: Survivabl Rsilin and Disrupion Tolran Nworking Univrsiy of Kansas Novmbr 1 2005

Oulin Moivaion: DTN rouing problm xampl Rouing wih zro knowldg (FC) Rouing wih parial knowldg (MEDEDEDLQEDAQ) Rouing wih compl knowldg (LP) Simulaion Rsuls Conclusion Novmbr 1 2005 Soshan Bali 2

DTN rouing problm (xampl) LAN ` DTN Gaway Inrn 6 hrs Congsion: - Bundl will b droppd Bundl will wai unil nx conac (whr will salli b?) ` 2 hrs Salli connciviy 1 hr afr arrival of bundl Novmbr 1 2005 Soshan Bali 3

Knowldg oracls Absrac knowldg oracls ar abl o answr qusions Conacs Summary Oracl: Aggrga saisics (.g. avg. waiing im unil nx conac for an dg) Conacs Oracl: Knowldg of all conacs for all dgs a all ims Quuing Oracl: Insananous quu lnghs of all nods Traffic Dmand Oracl: Prsn and fuur raffic dmand Novmbr 1 2005 Soshan Bali 4

Rouing wih zro knowldg Firs conac (FC) No knowldg oracls ar usd Forward along a randomly chosn dg (from availabl conacs) If no conacs availabl wai for firs conac Problms Mssag may no mak progrss owards h dsinaion Sourc Dsinaion Loops (spcially whn frqun conacs prsn among a s of nods) No provision o rou around congsion Improvmns Incorpora sns of rajcory o rou owards dsinaion Us pah vcor yp approach o avoid loops Novmbr 1 2005 Soshan Bali 5

Rouing wih parial knowldg MED ED EDLQ EDAQ On or mor of oracls: conacs summary conac and quuing Pah ha minimizs dlay incrass dlivry probabiliy Dijksra s algorihm finds minimum-cos pah Coss assignd o dgs o rflc simad dlay of mssag in aking ha dg Dijksra s algorihm hn finds h minimum-dlay pahs Algorihms diffr in h way coss ar assignd Coss may b Tim invarian (MED) Tim varying (ED EDLQ EDAQ) Limiaion of his mhod Only singl pah may b drivd Novmbr 1 2005 Soshan Bali 6

Inpu : G = Oupu : L 1: Q 2 : L Rouing wih parial knowldg Modifid Dijksra s algorihm ( V E) s T w( ) { V} [] s 0 L[] v v V _ s.._ v s _ Q {}_ do L _ u Q _ b _ h _ nod _ s _ L[] u = min x Q L[] x Q = Q {} u for _ ach _ dg _ E s _ = ( u v) _ do s if _ L[] v > ( L[] u + w( L[] u + T ))_ hn L[] v L[] u + w( L[] u + T ) 3: whil 4 : 5 : 6 : 7 : 8 : 9 : nd _ if 10 : nd _ for 11: nd _ whil Diffrn Novmbr 1 2005 Soshan Bali 7 T 1 2 [ u] = 1 2 L + u u u u u v ( L[ u] T ) = w + T + 1 T + 1 + T + 2 1+ 2 + 3 3 [ v] = 1 + 2 3 L +

Rouing wih parial knowldg Minimum Expcd Dlay (MED) Uss h conacs summary oracl Edg coss ar im-invarian (avrag im unil nx conac) Cos of an dg is avg. waiing im + prop. + ransmission dlay Rou of a mssag indpndn of im Problms No mchanism o rou around congsion Whn dirc conac bcoms availabl mssag sill wais for... pr-compud nx-hop Improvmns Modify pr-compud rou in ransi whn a conac bcoms availabl Novmbr 1 2005 Soshan Bali 8

Novmbr 1 2005 Soshan Bali 9 Rouing wih parial knowldg Tim-varying coss ED EDLQ and EDAQ assign im-varying coss L m is h mssag siz s is h nod assigning h cos ( ) ( ) s m w w ' = ( ) ( ) ( ) ( ) ( ) ( ) ( ) + = + = = " " min ' ' ' ' x s Q m dx x c s m d s m s m w m ( ) s Q ( ) m s '

Rouing wih parial knowldg Earlis Dlivry (ED) Conacs oracl is usd Quuing informaion is no usd Rous ar loop fr (Dijksra s algorihm) Rous drmind a sourc and fixd (sourc rouing) Mssags may b droppd du o congsion Will work good whn Quus in pah ar mpy Conac capaciy is high (quus mpid in ngligibl im) Will no work good whn ( s) = 0 Q Mssags in quu conac may nd bfor mssag is sn Disasrous bcaus conac in nx hop may b missd Coninuing on rou compud arlir may no b opimal Novmbr 1 2005 Soshan Bali 10

Rouing wih parial knowldg Earlis Dlivry wih Local Quuing (EDLQ) Conacs oracl is usd Local quu occupancy is akn ino accoun ( s) Q ( s* ) daa_ quud_ for a _ im if _ = = 0 ohrwis Q accouns only for quuing a ougoing dgs of currn nod Hlps rou around congsion a firs hop Unlik ED r-compu rou a vry hop Problm May rsul in rouing loops Improvmn Us pah vcors o avoid loops Problm Lik ED mssags migh g droppd bcaus of buffr ovrrun Novmbr 1 2005 Soshan Bali 11

Rouing wih parial knowldg Earlis Dlivry wih All Quus (EDAQ) Conacs and Quuing oracls usd Insananous quu sizs across nir opology a any poin of im ar usd ( s) = daa_ quud_ for a _ im a _ nod_ s Q Lik ED mssags ar sourc roud Problm Congsion losss may occur bcaus EDAQ is oblivious o buffr sizs Novmbr 1 2005 Soshan Bali 12

Rouing wih compl knowldg LP formulaion LP formulaion is discussd in papr Objciv funcion is o minimiz h avrag dlay Conacs Quuing and Traffic oracls ar usd Buffr sizs ar assumd o b known Evn for a vry simpl opology compuaion im is abou 8 minus.. on a 8 procssor P3 machin No fasibl o compu rous in ral-im for ralisic opologis Novmbr 1 2005 Soshan Bali 13

Simulaion Sup Scnario 1: Rouing o a rmo villag Dialup: 4Kbps (availabl 11AM o 6PM) LEO salli (posiion drmind by racking sofwar) Moorbik (2 hrs on way bandwidh o/from moorbik 1Mbps can sor up-o 128 MBys in USB sick) Mssags: villag o ciy 1KB rvrs dir. 10KB low load = 200 mssags pr day high load = 1000 mssags mssags gnrad for on day simulaion im = 2 days Novmbr 1 2005 Soshan Bali 14

Simulaion Rsuls Scnario 1: Rouing o a rmo villag MED rous all daa ovr dialup (bs avrag dlay im-invarian) ED: mos daa salli rs dialup (moorbik no usd: salli crossings lss han 2 hrs) FC prforms ok for simpl opology (all pahs lad o h ciy) Congsion awar schms us moorbik a high loads ED and MED rmain sam a low and high loads (no congsion awar) Boh ED and MED suffr bcaus of no bing congsion awar EDLQ and EDAQ avrag dlay prformanc is clos o LP Novmbr 1 2005 Soshan Bali 15

Simulaion Sup Scnario 2: a nwork of ciy buss Map of San Francisco usd for bus movmns of 20 buss Bus sop whn bus maks a urn o a nw sr Sop im 0 o 5 mins uniform disribud Bus spd 10-20 m/sc uniform disribud Buss communica whn in radio rang: dfaul = 100m Mssags gnrad for 12 hrs 20 random sourc dsinaion pairs chosn in ach hour Each sourc bus snds 200 mssags o is dsinaion in 1 hr Mssags injcd simulanously a a randomly chosn im in h 1 hr inrval This rprsns bursy raffic Load=Traffic dmand/conac volum Evn wih a load lss han 1 nwork may b congsd (mssag has o ravrs mulipl hops conacs may b prsn bu no raffic o uiliz hm) Load is changd by changing h conac volum (chang conac bandwidh or chang conac duraion radio rang) Dfaul sorag capaciy=100mb dfaul link bandwidh =100Kbps Novmbr 1 2005 Soshan Bali 16

Simulaion Rsuls Scnario 2: a nwork of ciy buss Chang load: bandwidh basd Low loads: dlay dos no chang problm is conacs no bandwidh Highr loads: hr is congsion EDAQ EDLQ prform br Vry high load: mos no dlivrd Avrag dlay dcrass wih incras in radio rang High radio rang buss in conac highr prcn of im Diffrnc in algorihms mor pronouncd whn connciviy is inrmin Novmbr 1 2005 Soshan Bali 17

Conclusion EDEDLQEDAQ ouprform FCMED: boh dlay and dlivry raio ED prforms wors han EDLQEDAQ: congsion miigaion Prformanc diffrnc mor pronouncd whn conacs inrmin EDLQ and EDAQ prformanc comparabl LP canno compu rous in ral-im Papr assums ha knowldg oracls ar prsn Ralizing knowldg oracls in ral-world is a big problm (wih frqun disconncs and high dlays) Algorihms may b usful in spcial cas scnarios whr conacs can b prdicd Novmbr 1 2005 Soshan Bali 18