ECEN4533 Data CommunicationsLecture #21 25 February 2013Dr. George Scheets

ECEN4533 Data CommunicationsLecture #21 25 February 2013Dr. George Scheets

Read 11.4Read 11.4 Problems: Chapter 11.2, 4, & 5Problems: Chapter 11.2, 4, & 5 Quiz #2, 25 March (Live) Quiz #2, 25 March (Live) << 1 April (DL) 1 April (DL)

ECEN4533 Data CommunicationsLecture #22 27 February 2013Dr. George Scheets

ECEN4533 Data CommunicationsLecture #22 27 February 2013Dr. George Scheets

Read 11.5Read 11.5 Corrected ExamsCorrected Exams

Due 6 March (Live)Due 6 March (Live) One week after return (DL)One week after return (DL)

Scan Design Problem #2Scan Design Problem #2

ECEN4533 Data CommunicationsLecture #23 1 March 2013Dr. George Scheets

ECEN4533 Data CommunicationsLecture #23 1 March 2013Dr. George Scheets

Read 11.6.1, 11.7Read 11.6.1, 11.7 Problems Web 13-15Problems Web 13-15 Corrected ExamsCorrected Exams

Due 6 March (Live)Due 6 March (Live) One week after return (DL)One week after return (DL)

Design #2Design #2 Due 15 March (Live)Due 15 March (Live) Due 22 March (DL)Due 22 March (DL)

RedNeckNetLow BidRedNeckNetLow Bid

Cost = $12, 643.40 per monthCost = $12, 643.40 per month

Promoted to MegaMoron Senior EngineerPromoted to MegaMoron Senior Engineer Tsega DebeleTsega Debele

Design #2 Improved RedNeckNetDesign #2 Improved RedNeckNet Combined Traffic Options (Pick Combined Traffic Options (Pick OneOne):):

(1) Standard Internet (FIFO)(1) Standard Internet (FIFO) (2) QoS Enabled Internet(2) QoS Enabled Internet (3) ATM: VBR Video, UBR Data(3) ATM: VBR Video, UBR Data

Redundancy: Two ConnectivityRedundancy: Two Connectivity AnalysisAnalysis

Calculate Single Hop DelaysCalculate Single Hop Delays Queuing + Propagation DelaysQueuing + Propagation Delays

Multiple Hop Delays = sum of single HopsMultiple Hop Delays = sum of single Hops Make sure you account for the overhead!!!Make sure you account for the overhead!!!

Traffic Matrix is application trafficTraffic Matrix is application traffic

Design Comments:Design Comments: Link CostsLink Costs

Proportional to distanceProportional to distance1.01.0traffictraffic0.00.0??Centralized Star is best.Centralized Star is best.

Proportional to distanceProportional to distance0.00.0traffictraffic1.01.0??Full Mesh is Best.Full Mesh is Best. Ours is proportional to distanceOurs is proportional to distance0.220.22traffictraffic0.770.77

Otherwise pay attention to traffic matrixOtherwise pay attention to traffic matrixHigh traffic node?High traffic node?Generally best to have many links.Generally best to have many links.

Classical Queuing TheoryClassical Queuing Theory

M/x/xM/x/x Exponentially Distributed IATExponentially Distributed IAT

M/G/1, M/M/1, M/D/1 M/G/1, M/M/1, M/D/1 Single server, various packet distributionsSingle server, various packet distributions

M/M/1 with prioritiesM/M/1 with priorities M/M/aM/M/a

Several (a) output serversSeveral (a) output servers

Real vs Artificial TraceReal vs Artificial Trace

10 SecondsReal Traffic 10 Seconds

Artificial M/M/1 Traffic

Source: Willinger et al, "Self-Similarity through High Variability", IEEE/ACM Transactions on Networking, February 1997.

Real vs Artificial TraceReal vs Artificial Trace

100 SecondsReal Traffic 100 Seconds

Artificial M/M/1 Traffic

Real vs Artificial TraceReal vs Artificial Trace

16.7 MinutesReal Traffic 16.7 Minutes

Artificial M/M/1 Traffic

Real vs Artificial TraceReal vs Artificial Trace

167 MinutesReal Traffic 167 Minutes

Artificial M/M/1 Traffic

Real vs Artificial TraceReal vs Artificial Trace

27.78 HoursReal Traffic 27.78 Hours

Artificial M/M/1 Traffic

Self Similar BehaviorSelf Similar Behavior

Fall 2002 FinalFall 2002 Final 'Average' based on 1 test chosen at random'Average' based on 1 test chosen at random

126.00 out of 150126.00 out of 150 One point average XOne point average X11

'Average' based on 10 tests chosen randomly'Average' based on 10 tests chosen randomly109.44 out of 150109.44 out of 150 Ten point average XTen point average X1010

Actual Midterm AverageActual Midterm Average106.85 out of 150106.85 out of 150

Harold Edwin Hurst (1880-1978)Harold Edwin Hurst (1880-1978)

British Hydrologist

Found Long Term Dependence in Reservoir Storage.

One way to estimate HOne way to estimate H Compute Variance of original trace, XCompute Variance of original trace, X Merge 10 consecutive points into new point YMerge 10 consecutive points into new point Y

Y = 10 point average of X Y = 10 point average of X Compute variance of this new traceCompute variance of this new trace

Merge 10 consecutive points of Y into new point ZMerge 10 consecutive points of Y into new point ZZ = 100 point average of XZ = 100 point average of XCompute variance of this new traceCompute variance of this new trace

How is variance decreasing?How is variance decreasing? By factor of N? Not self-similar, H = 0.5By factor of N? Not self-similar, H = 0.5 By factor of NBy factor of N2(1-H)2(1-H)? Evidence of self-similarity? Evidence of self-similarity

Classical Queuing TheoryVs. Self Similar

Classical Queuing TheoryVs. Self Similar

0% 100%Offered Load

M/D/1M/M/1AverageDelay

SelfSimilar

Traffic Rate over timeTraffic Rate over time

Source: "Dimensioning Network Links", IEEE Network Magazine, April 2009

Carrying Capacity% of Line Speed carrying Application TrafficCarrying Capacity% of Line Speed carrying Application Traffic

Line Speed

Load (63% Active) 37% Idle

84% Traffic 16% Overhead

Carrying Capacity = Traffic Moved/Line Speed = 84% * 63% = 53%

Carrying CapacityCarrying Capacity Estimate Packet/Cell overhead Estimate Packet/Cell overhead Estimate Traffic Characteristics Estimate Traffic Characteristics

Classical?Classical? Self-Similar? What is Hurst Parameter?Self-Similar? What is Hurst Parameter?

Estimate allowable trunk loadEstimate allowable trunk load Delay constrained? Delay constrained? (low speed tendency)(low speed tendency) Buffer constrained? Buffer constrained? (high speed tendency)(high speed tendency)

Calculate Carrying Capacity =Calculate Carrying Capacity =(% Trunk Load)(% of Traffic in packet/cell)(% Trunk Load)(% of Traffic in packet/cell)

Low Speed LinksLikely Delay Constrained

Plenty of Memory Available

Low Speed LinksLikely Delay Constrained

Plenty of Memory Available

0% 100%

Offered Load

AverageDelay

50%

TargetAverage

TargetMaximum

PDF

High Speed LinksLikely Memory Constrained

Plenty of Time Available

High Speed LinksLikely Memory Constrained

Plenty of Time Available

0% 100%

Offered Load

50%

MemoryLimit

QueueSize

TargetAverage

Circuit Switch TDM Trunking(Leased Line Telephone Network)

Circuit Switch TDM Trunking(Leased Line Telephone Network)

TDMSwitch

SONET OC-N

Fixed Rate Traffic

Bursty Data Traffic

Assumptions: Fixed Rate Traffic gets fixed number of time slots.(N Bytes every 1/8000th second). Bursty Data Traffic channels get fixed number oftime slots based on peak (line) input rates.

Carrier Leased Line NetworkCarrier Leased Line Network

Leased Line ‘Cloud’Trunk capacity shared via TDM & Circuit Switching

Cross-ConnectCross-Connect

TrunksLeased Line

Packet Switch StatMux Trunking(Pure Internet Model)

Packet Switch StatMux Trunking(Pure Internet Model)

Router

SONET OC-N

Fixed Rate Traffic

Bursty Data Traffic

Assumptions: All traffic is packetized & Statistically Multiplexed onto the trunk BW.

Internet Service Provider BackboneInternet Service Provider Backbone

Router

TrunksLeased Line

ISP ‘Cloud’Trunk capacity shared via StatMux & Packet Switching

Cell Switch StatMux Trunking(ATM Model)

Cell Switch StatMux Trunking(ATM Model)

ATMSwitch

SONET OC-N

Fixed Rate Traffic

Bursty Data Traffic

Assumptions: Fixed rate traffic moved over CBR VC's. Gets reserved bandwidth and near-TDM like service. Data or Variable Rate Traffic is StatMuxed onto the trunk bandwidth that’s not reserved for CBR.

ATM BackboneATM Backbone

ATM ‘Cloud’ Trunks use StatMux/TDM & Cell Switching

ATM SwitchATM Switch

TrunksLeased Line

Switched Network Carrying CapacitiesHigh Speed Trunk (OC-3)

Switched Network Carrying CapacitiesHigh Speed Trunk (OC-3)

0% Bursty 100% Bursty100% Fixed Rate 0% Fixed Rate

Offered Mix

Carrying Capacity

Circuit Switch TDM

Packet Switch StatMux

Cell Switch StatMux

Traffic GrowthTraffic Growth

Voice

Data

time

70’s & 80’s Voice Dominates70’s & 80’s Voice Dominates

Voice

Data

time70’s & 80’s

Switched Network Carrying CapacitiesHigh Speed Trunks

Switched Network Carrying CapacitiesHigh Speed Trunks

Carrying Capacity

Circuit Switch TDM

0% Bursty 100% Bursty100% Fixed Rate 0% Fixed Rate

OfferedTraffic Mix

Turn of the CenturyA Mixed Traffic EnvironmentTurn of the CenturyA Mixed Traffic Environment

Voice

Data

time2000

Switched Network Carrying CapacitiesHigh Speed Trunks

Switched Network Carrying CapacitiesHigh Speed Trunks

Carrying Capacity

Cell Switch StatMux

0% Bursty 100% Bursty100% Fixed Rate 0% Fixed Rate

OfferedTraffic Mix

Today, Data DominatesToday, Data Dominates

Voice

Data

time

2013

Switched Network Carrying CapacitiesHigh Speed Trunks

Switched Network Carrying CapacitiesHigh Speed Trunks

Carrying Capacity

Packet Switch StatMux

0% Bursty 100% Bursty100% Fixed Rate 0% Fixed Rate

OfferedTraffic Mix

The Big Unknown...

What impact will Video have?

The Big Unknown...

What impact will Video have?

If Video becomes dominant,If Video becomes dominant,is a packet switched statmux is a packet switched statmux network best?network best?

ISO OSI Seven Layer ModelISO OSI Seven Layer Model

Layer 7 ApplicationLayer 7 Application Word PerfectWord Perfect Layer 6 Presentation Layer 6 Presentation Windows APIWindows API Layer 5 SessionLayer 5 Session TCP, WindowsTCP, Windows Layer 4 TransportLayer 4 Transport TCP, WindowsTCP, Windows Layer 3 NetworkLayer 3 Network IP, WindowsIP, Windows Layer 2 Data LinkLayer 2 Data Link PC NICPC NIC Layer 1 Physical Layer 1 Physical PC NICPC NIC