aqm performance for voip

Performance Analysis of AQM for VOIP Application

MakkawiAbdelsalmMohammed KhairFuture university ,249915001523, khairmakkawi@gmail.com 

  Dr. Mohammed Hussain  Supervisor

• VOIP • Problems • Objectives• AQM

• RED• ARED• GRED• WRED

• Scenario 1• Scenario 2 • Results • Conclusion• Future work

Overview

VOIP

• Voice Over Internet Protocol• Is the technology that uses the Internet Protocol to transmit voice conversation over a data network.

• The primary advantages of moving voice over a data network are increased efficiency and decrease cost.

• Delay• Jitter• Packet loss

VoIP QoS Issue

Enterprise: Private Branch Exchange (PBX)

7043

7040

7041

7042

External line

Telephoneswitch

Private BranchExchange

212-8538080

Anotherswitch

Corporate/Campus

InternetCorporate/Campus LAN

Post-divestiture phenomenon...

Enterprise VoIP: Yesterday’s networksCircuit Switched Networks (Voice)

Packet Switched Networks (IP)

PBXPBX

COCO

CO

Router

RouterRouter

Router

Router

Headquarters Branch Offices

Enterprise VoIP: Today’s networksToll by-pass

Circuit Switched Networks (Voice)

Packet Switched Networks (IP)

PBXPBX

COCO

CO

Router

RouterRouterRouter

Router

Headquarters Branch Offices

Enterprise VoIP: Tomorrow’s networksUnified/Converged Networks

Unified Networks (Voice over IP)

Router

RouterRouter

Router

Router

COCO

Legacy PSTN

Headquarters Branch Offices

ProblemsThe main problem of this work is that voice is highly sensitive for delay and packet loss

in network (Cisco Systems, 1992–2008).

It is crucial to have a proper QoS mechanism to guarantee minimum delay and packet

loss

Therefore, it is the purpose of this research to determine how does the AQM mechanisms

deal with the services such as VoIP. Furthermore, it will try to investigate what are the

most powerful mechanisms that can provide the best results to meet the objective of the

Quality of Services (QoS).

Objectives

It is the objective of this research to compare the performance of four mechanism of AQM in terms of packet delay,

packet loss, variation of delay and throughput.

Specifically it would like to achieve the following: 1. Determine the performance of RED (Random Early Detection) in terms of:

a) Packet delayb) Packet lossc) Variation of packet delayd) Throughput

2. Determine the performance of ARED (Adaptive Random Early Detection) in terms of a) Packet delayb) Packet lossc) Variation of packet delayd) Throughput

Objectives3. Determine the performance of GRED (Gentle Random Early Detection) in terms of  

3. Packet delay4. Packet loss5. Variation of packet delay6. Throughput

4. Determine the performance of WRED (Weighted Random Early Detection) in terms of  a) Packet delayb)Packet lossc) Variation of packet delayd)Throughput

12

Active Queue Management

Queue

SinkOutbound LinkRouterInbound Link

SinkACK…

ACK…

Queue

SinkOutbound LinkRouterInbound Link

SinkACK…

ACK…

Queue

SinkOutbound LinkRouterInbound Link

SinkACK…

Drop!!!

Queue

SinkOutbound LinkRouterInbound Link

Sink

Queue

SinkOutbound LinkRouterInbound Link

Sink

AQM

Congestion

Congestion Notification…

ACK…

Queue

SinkOutbound LinkRouterInbound Link

Sink

AQM

Advantages• Reduce packet losses  (due to queue overflow)• Reduce queuing delay

scr2

scr1

Let us See How This Mechanisms Works ….

RED (Random Early Detection)

PP

PP

PP

PP

PP

PP

RED MECHANISM

minmax

PP

PP

PP

PP

PP

PP

RED MECHANISM

minmax

P

PP

Case1 : packets < min threshold = no drop

PP

PP

PP

PP

PP

PP

RED MECHANISM

minmax

P

PP

P

PP

Case2 : min threshold< packets< max threshold = drop with probability

PP

PP

PP

PP

PP

PP

RED MECHANISM

minmax

P

PP

PP

P

PP

Case2 : min threshold< packets< max threshold = increase the drop

PP

PP

PP

PP

PP

PP

RED MECHANISM

minmax

P

PP

PP

PP

PP

P

PP

PPP

Case3 : packets > max threshold = drop all incoming packets

What About ARED ?!!

(Adaptive Random Early Detection)

PP

PP

PP

PP

PP

PP

ARED MECHANISM

minmax

PP

PP

PP

PP

PP

PP

ARED MECHANISM

min

P

PP

max

Case1 : packets > min threshold = adapt the queue (decreas max threshold)

PP

PP

PP

PP

PP

PP

ARED MECHANISM

min

P

PP

P

PP

max

Case1 : packets < min threshold = drop with probability

PP

PP

PP

PP

PP

PP

ARED MECHANISM

minmax

P

PP

PP

P

PP

Case2 : packets < min threshold = adapt the queue (move min threshold)

P

P

PP

PP

PP

PP

PP

ARED MECHANISM

minmax

P

PP

PP

P

PP

P

PPP

PP

P

PP

max

PPP

Case3: packets > max threshold = adapt the queue (move max threshold)

Drop all incoming packets

We still have GRED

(Gentle Random Early Detection)

P

P

PP

PP

PP

PP

PP

GRED MECHANISM

minmax

P

PP

P

PP

P

PP

PP

PP

PP

P

PP

minmaxDouble max

P

PP

P

PP

P

PP

P

PP

P

PP

P

PP

PPP

Increase the length of the queue by doubling the maximum threshold

Finally we have WRED

(Weighted Random Early Detection)

PP

PP

PP

PP

PP

PP

WRED MECHANISM

minmax

PPP

PPP

PPP

PPP

PPP

PPP

P High priority P Mid priority P Low priority

Deal with the data according to it’s priority

OPNET Object Hierarchy

C/C++ Code

Process

Node

Subnet

Scenario

Project

First Scenario

First scenario parameters

• This scenario based on changing the number of users while the other parameters remain constant

• The below table shows the simulation parameters that used in this scenario

Simulation parameters

Number of User Variable from 5-300

Data forwarding rate (packets/second) 500000

Maximum queue size 100

Minimum threshold 10

Maximum threshold 20

Mark probability denominator 0.1

Simulation duration 1 hour

Encoder scheme G.711 (silence)

Voice frame per packet 5

Dumbbell Topology

10 users results

50 Users Topology

50 Users Results

300 Users Topology

300 Users Results

Second Scenario

Second scenario parameters

• This scenario based on changing the number of users while the other parameters remain constant

• The below table shows the simulation parameters that used in this scenario

Simulation parametersNumber of User 10 ,50,300

Data forwarding rate (packets/second) 500000

Maximum queue size Variable 200,300,500,1000

Minimum threshold 20

Maximum threshold 60

Mark probability denominator 0.1

Simulation duration 1 hour

Encoder scheme G.711 (silence)

Voice frame per packet 5

Maximum queue size(packets)=200 (jitter)

10 users  50 users  300 users 

Maximum queue size(packets)=200 (drop)

Router A 50 users  Router A 300 users Router A 10 users 

Maximum queue size(packets)=200 (drop)

Router B 10 users  Router B 50 users  Router B 300 users 

Maximum queue size(packets)=200 (delay)

10 users  50 users  300 users 

Maximum queue size(packets)=1000 (Jitter)

10 users  50 users  300 users 

Maximum queue size(packets)=1000 (drop)

Router A 10 users  Router A 50 users  Router A 300 users 

Maximum queue size(packets)=1000 (drop)

Router B 10 users  Router B 50 users  Router B 300 users 

Maximum queue size(packets)=1000 (delay)

10 users  300 users 50 users 

Conclusion• This study evaluated the performance of VoIP over Ethernet WAN by applying AQM mechanism

using OPNET simulation tool. • A comparison has been conducted between four of (AQM Active Queue Management) methods

RED, ARED, GRED and WRED This comparison aimed to identify which method offers more satisfactory performance measures for application like VOIP.

• A decision which method offers more satisfactory performance measure results is only made depending on varying the number of users and number of maximum queue size in the routers

• It has been found that the configuration of the RED algorithm and GRED is not axiomatic. They must consider many situations such as network traffic, the traffic type, nature and duration of possible bursts, delays in the network, etc.

• It has been seen that in general, the ARED and GRED algorithms performs better, achieving a lower discard rate and lower overall delay. One problem that has been is the average queue size parameter tends to oscillate, especially given configuration changes and wq the minimum threshold, so care must be taken to configure.

Future work • This study used the comparison according to one type of traffic, it recommend that to use some

other traffic such as Http or video to see how this mechanism will deal with the VoIP packets while there is another types of packet.

• Other parameters could be considered such as different encoder scheme and number of voice frame per packet.

Thank you