A Bandwidth Estimation Method for IP Version 6 Networks

A Bandwidth Estimation Method for IP Version 6 Networks

Marshall Crocker

Department of Electrical and Computer Engineering

Mississippi State University

October 13, 2006

Page 2 of 37MS Thesis Presentation

Outline

• Introduction to Bandwidth and Estimation

• Motivation

• IPv4 Estimation Techniques

• IPv6 Overview and Estimation Technique

• Simulation Experimentation

• Conclusions and Future Work


What is Bandwidth?

• Important characteristic of data networks How much data How fast

• Determined by two primary properties Physical Link Capacity Infrastructure Utilization

Data?

100 Mb/s10 Mb/s

% Load


What is Bandwidth Estimation?

• Nodes attempt to estimate network bandwidth Determine minimum physical capacity called Bottleneck Bandwidth Determine unused capacity called Available Bandwidth

• Estimations are used in many different ways

• Many different techniques for performing estimations

100 Mb/s10 Mb/s

90% Load


Sending rate less than

BWE?

Applications of BWE

• End-to-end flow control

Increase Rate

Decrease Rate

yes

no


Applications of BWE

• Server selection for downloads and streaming media

• Peer-to-peer selection Connect to peers with most bandwidth

• Traffic Engineering Adjust routing/switching for optimal operation

• Capacity Provisioning Increase/decrease capacity as needed

10

5

8


Motivation

• BWE valuable for a number of applications

• IPv4 techniques Suffer from various flaws Limited due to nature of the network No single technique suitable for all applications Evolving network technologies affect theories

• IPv6 offers framework for improved estimation technique Efficient Flexible Accurate Simple


IPv4 Estimation Techniques

• All measurements are passive

• Examine how network delivers data Packet spacing Packet delay Packet dispersion Statistical


IPv4 Estimation Techniques

• Each suffer from one or more High level of complexity Poor efficiency Limited accuracy Application specificity

• Each method is susceptible to one or more Network load Cross-traffic Packet-size variability Probing packet size Train length Cross-traffic routing


Incorrect IPv4 Bandwidth Estimation Scenario


IP Version 6

• Next generation Internet Protocol

• Improves on IPv4 Expanded addressing from 32 bits to 128 Simplified header Improved extension and option support

• Extension support provides framework for improved bwe technique


Data

….

IPv6 Header

Ver 6 Traffic Class Flow Label

32 bits

Payload Length Next Hdr. Hop Limit

Source Address

Destination Address

Extensions


IPv6 Extensions

• Several different extensions Routing Fragmentation Destination options Authentication Security Hop-by-hop

• Examined by every hop • Provides instructions for each hop• Only two options currently defined

Jumbo payload Router alert


Proposed Hop-by-Hop Options

• Traceroute Each hop inserts address Record forward/backward path Not accepted by IETF

• Connection Status Investigation (CSI) Request statistics/attributes for each hop IP address Bandwidth Type Number of transmitted/received bytes/packets Number of errors


IPv6 Timestamp Option

• CSI would have been extremely useful

• Rejected by IETF due to complexity, security, and proprietary concerns

• A timestamp option was defined for IPv4 but had limited use

• An IPv6 timestamp option has much more potential including bandwidth estimation


IPv4 Timestamp Deficiencies

• IPv4 timestamp option limited in usefulness

Can only hold timestamps for up to 9 hops without addresses

Room to hold 4 hops with addresses

No standard for defining timestamp format

IPv4 routers services packets with options slower


IPv6 Timestamp

• IPv6 timestamp properties

Enough room to hold timestamp records for every hop

Predefined timestamp format

Timestamp at incoming and/or outgoing interfaces


IPv6 Timestamp Format

Next Hdr. Hdr. Ext Len Option Type Option Data Len

32 bits

Record Count TS Type Hop Limit BaseRes IfOptR

Identifier Reserved

Upper Part of IPv6 Address

Lower Part of IPv6 Address

Hop Number CounterLk TypeI/FResolutionG

TimestampFmt


IPv6 Timestamp BWE

• Define bandwidth as number of transmitted bits per unit time

• Expand to include start and end transmit times

• Use start/end transmit times of packet and packet size to calculate capacity

• Send two timestamp packets back-to-back

• Timestamp of first packet and timestamp of second packet = t1 and t2

• Size of first packet and link layer size used in final calculation

t

bC

12 tt

bC


Bottleneck Bandwidth Estimation

LeadTail

Router

TS = 10TS = 15

1015

SizeLeadPacketC


Bottleneck Bandwidth Estimation

• Relies on back-to-back queuing

• Count field in TS record ensures back-to-back

• Smaller tail packet helps back-to-back queuing


1020

SizeLeadPacketC

Available Bandwidth Estimation

LeadTail

Router

TS = 10TS = 20

CT


Available Bandwidth Estimation

• Relies on cross traffic to introduce packet separation

• Constantly changing value

• Applications must send estimations frequently


Simulation Experimentation

• Simulation experiments used to compare and evaluate IPv6 Timestamp method

• Measured against comparable IPv4 method called the cartouche method

• Cartouche method uses packet trains and examines packet spacing to estimate BW


Simulation Setup


Simulation Setup

Parameter Values

Number of Cross Traffic Flows 8, 16, 24, 32, 40, 48

Cross Traffic Rates 100 Kbps, 1 Mbps

Average Cross Traffic Size 200 Bytes

Leading Probe Packet Size 1500 Bytes

Bottleneck Rates (Mbps) 10, 20, 30, 40, 50,60, 70, 80, 90, 100

Hop Rates 100 Mbps


Estimation Method Parameters

Scenario CT Rate Tailgating Packet Size (Bytes)

1 100 Kbps 212

2 100 Kbps 76

3 1 Mbps 212

4 1 Mbps 76

Scenario CT Rate Tailgating Packet Size (Bytes)

Cartouche Length (r)

1 100 Kbps 40 2

2 100 Kbps 40 3

3 1 Mbps 40 2

4 1 Mbps 40 3


IPv6 Estimation Results

Scenario 1 Scenario 4


Cartouche Estimation Results

Scenario 1 Scenario 4


IPv6 Estimation Frequency

IPv6 Bandwidth Measurements 100 Kbps Cross Traffic Flows

0

100

200

300

400

500

600

700

0 10 20 30 40 50 60 70 80 90 100

Measured Bandwidth (Mbps)

Fre

qu

ency


IPv6 Estimation Frequency

IPv6 Bandiwdth Measurements 1 Mbps Cross Traffic Flows

0

100

200

300

400

500

600

0 10 20 30 40 50 60 70 80 90 100


Fre

qu

en

cy


Cartouche Estimation Frequency

Cartouche Bandwidth Measurements 100 Kbps Cross Traffic Flows

0

50

100

150

200

250

300

350

400

450

500

0 10 20 30 40 50 60 70 80 90 100


Fre

qu

en

cy



Cartouche Bandwidth Measurements for 1 Mbps Cross Traffic Flows

0

20

40

60

80

100

120

140

0 10 20 30 40 50 60 70 80 90 100


Fre

qu

enc

y



Cartouche Bandwidth Measurements 100 Kbps Cross Traffic Flows Bottleneck = 90 Mbps

0

10

20

30

40

50

60

70

80

0 10 20 30 40 50 60 70 80 90 100


Fre

qu

en

cy



Cartouche Bandwidth Measurements for 1 Mbps Cross Traffic FlowsBottleneck = 90 Mbps

0

2

4

6

8

10

12

14

16

18

0 10 20 30 40 50 60 70 80 90 100


Fre

qu

en

cy


Conclusions

• Presented IPv6 bandwidth estimation using timestamp hop-by-hop option

• Advantageous over existing methods Efficient Simple Flexible Accurate

• IPv4 bandwidth estimations are limited due to the nature of the network

• Outperforms comparable IPv4 Technique


Future Work

• Extended simulation models Diverse network properties and conditions Additional hardware and communications models Additional host and network models

• Real world implementation

• Development of network control techniques, protocols and applications such as a “Cognizant” version of TCP Aware of network Intelligently respond to network and conditions Fairly use network resources

A Bandwidth Estimation Method for IP Version 6 Networks

Documents

ms thesis presentationwhat

bandwidth estimation

different techniques

peer selection

different ways

improved bwe techniquepage

theories ipv6

streaming media peer