Example Applications needing Advanced Services

Preview:

DESCRIPTION

Example Applications needing Advanced Services. Campus Focused Workshop on Advanced Networks Atlanta, GA. Voice over IP Environment for Research (VIPER). Chakravarthy K Sannedhi Electrical & Computer Engineering. VoIP: Benefits. Data traffic growing rapidly - PowerPoint PPT Presentation

Citation preview

Example Applications needing Example Applications needing Advanced ServicesAdvanced Services

Campus Focused Workshop on Advanced Networks

Atlanta, GA

Voice over IP Environment for Voice over IP Environment for Research (VIPER)Research (VIPER)

Chakravarthy K Sannedhi

Electrical & Computer Engineering

VoIP: BenefitsVoIP: Benefits

Data traffic growing rapidly– Multiple parallel networks are expensive– VoIP Enables convergence of voice & data

networks – Low-cost and flat-rate pricing possible

IP is compatible with most modern network technologies & topologies

VoIP: IssuesVoIP: Issues

Delay– too much can cause “real time” voice interaction to

become useless

Jitter– small amounts can be “fixed” by jitter buffers, but end-

to-end delay suffers

Packet loss– creates “big gaps” in received (reconstructed) voice

VoIP: QoS approachesVoIP: QoS approaches

Bandwidth allocation– Necessary, but not sufficient

Prioritization of the voice traffic– Necessary, but not sufficient– Different techniques have different effects on voice as

well as “other traffic” QoS Techniques

– DiffServ– IntServ– Traffic Shaping

VoIP: Quality TestingVoIP: Quality Testing

Objective Testing– Necessary, but not terribly useful– Objective measures often correlate poorly with human perception

(even when perceptual models are used) Subjective Testing

– Necessary, but prone to error and interpretation– Time consuming, particularly in context of network-dependent

error mechanisms General idea … VIPER

– Automated environment to configure various network-based parameters which affect voice QoS

– Enables collecting of subjective test data

VIPER ArchitectureVIPER Architecture

VIPER ArchitectureVIPER Architecture

Web Interface

Call Generator Test taker’s data collector

Noise Generator Script Loader

Voice Database

Script Database

MySQL Database

VIPER ArchitectureVIPER Architecture

QoS Techniques TestedQoS Techniques Tested

IntServ (RSVP)DiffServ (EF and AF)Label Switching (MPLS)Traffic Shaping (CAR, GTS, etc.)

Best EffortBest Effort

No QoSFirst In First OutStill the voice is marked with EF1.1

RSVPRSVP

IntServ TechniqueSender sends the PATH message which

contains TSpecReceiver sends RESV which includes

Flowspec75% of the bandwidth to voice1.4

Weighted Fair QueuingWeighted Fair Queuing

Schedules interactive traffic to the front of the queue

Applies weights to identified traffic flowsShares the remaining bandwidth between

the high bandwidth flows3.6

Custom QueuingCustom Queuing

Services the traffic in round robin basisVoice was given maximum queue limit and

maximum byte count1.2

Priority QueuingPriority Queuing

Suitable for interfaces with less than 2.048 Mbps bandwidth

Voice is placed in the High priority queueInjustice to traffics that are other than in

High priority queue1.2

Class Based WFQClass Based WFQ

Traffic is placed in different classesSimultaneous handling of the traffic10 % of bandwidth to voice4.4

CBWFQ with LLQCBWFQ with LLQ

Brings strict priority queuing to CBWFQPreferential treatment for the voiceNot effective on Frame Relay networks3.6

Committed Access RateCommitted Access Rate

Traffic Shaping techniqueVoice packets are given the nice burst range

with a good amount of toleranceLowers the Jitter1.4

IP RTP Priority with WFQIP RTP Priority with WFQ

Useful for slow speed links with speeds less than 1.544 Mbps

Voice packets are identified by the UDP port range

Voice was given 60 Kbps of bandwidth along with the application of fair queuing

3.9

IP RTP Priority with RSVPIP RTP Priority with RSVP

Voice is identified by its UDP port range75% of the bandwidth to the voice 1.5

VIPER pMOS resultsVIPER pMOS resultsQoS (in extreme congestion) PMOS Missed Calls

(%)Confidence

(95%)Best Effort 1.1 26.9 1.1 0.14

WFQ 3.6 0 3.6 0.27

RSVP only 1.4 7.7 1.4 0.28

IPRTP + EF 1.4 19.2 1.4 0.28

IPRTP + WFQ 3.9 0 3.9 0.27

RSVP + IPRTP 1.5 15.4 1.5 0.28

CQ + EF 1.2 7.7 1.2 0.19

PQ + EF 1.2 19.21.2 o.19

CBWFQ + LLQ + EF 3.6 0 3.6 0.27

CBWFQ + EF 4.4 0 4.4 0.22

CAR + EF 1.4 0 1.4 0.36

AcknowledgementsAcknowledgements

Jill Gemmill – Assistant Director, Department of Academic Computing, UAB

Stan McClellan – Associate Professor, Electrical & Computer Engineering. UAB

ReferencesReferences

Red Hat Linux - http://www.redhat.comIperf - http://dast.nlanr.net/Projects/IperfMySQL - http://www.mysql.comPHP - http://www.php.netCisco - http://www.cisco.comExpect - http://expect.nist.govVgetty - http://alpha.greenie.net/vgetty

Web Resources for the ProjectWeb Resources for the Project

http://www.dpo.uab.edu/~kalyan/proreport.html

Recommended