Capacity Estimation of VOIP Channels on Wireless Networks

1
Capacity Estimation of VOIP Channels on Wireless
Networks

• Baek, Seung Jun
• Cui, Ningzhuo
• Ogale, Vinit
• Park, Ryoungsun
• Patel, Tanmay

2
Project Overview

• Project Goal
• Estimation of channel capacity with VoIP
  applications over wireless networks(802.11b)
• Measurement of channel characteristics (delay,
  jitter, packet loss) contributing to QOS for
  varying number of nodes

3
VoIP
Introduction
Network (Internet)
Voice
Voice Encoder
Voice Decoder
4
VoIP
Introduction

• Economical
• Convergence of wired and wireless data networks
  with PSTN and cellular networks
• Issues with VoIP over Wireless Links
• Is the QOS comparable to traditional PSTN ?
• How many simultaneous users can the network
  support ?

5
QoS

• What is QoS
• A metric which tries to measure the satisfaction
  level of the user
• Dependent on application
• Mean Opinion Score (MOS)
• A score of 1 to 5

6
Voice Samples

• MOS 4.5
• MOS 3.7
• MOS 3.2
• MOS 2.5

7
Factors Affecting QoS

• Throughput
• Delay
• Packet losses
• Jitter

8
Approach

• Measurement of capacity using the NetIQ Chariot
  network analysis tool
• Validation and extension of measurements by ns2
  simulations of the wireless network

9
Capacity Measurements

• Capacity
• Small scale
• Interference limited
• Determined by acceptable call quality
• Call quality
• Latency, jitter, packet loss
• MOS (Mean Opinion scores)

10
Chariot with VoIP Test Module

• Measures end-to-end performance
• Tests with real application loads
• Tests the capacity of the network
• Overall voice quality (MOS)

11
Measurement Setup
Network (Internet)
12
MOS, Throughput, and Delay
13
VoIP Capacity Measurement
14
Network Simulator NS2

• NS version2 is a discrete-event driven and
  object-oriented network simulator
• Type of simulation continuous, discrete event,
  and combined
• NS is an Object-oriented Tcl(OTcl) script
  interpreter that has a simulation event scheduler
  and network component object libraries, and
  network setup module libraries.
• The current NS version is written in C and Tcl.

15
NS2 Users View

• This is an overview of NS2
• Tcl commands translates into
• A series of object creations
• Plumbing of these objects

16
NS2 Simulation Parameters

• Network Interfaces - The model approximates the
  DSSS radio interface.
• The mac layer 802.11b
• Radio Propagation model The two ray ground
  bounce model.
• Antenna Omni-directional antenna having a unity
  gain.

17
Simulation Setup(1)

• Random Call Model
• Poisson Call Arrivals
• Exponential Call Holding Time
• Generates a file Call Profile

Call_Profile.cfg Format s1 d1 start end s2 d2
start end
18
Simulation Setup(2)

• 1520 VOIP Clients on 200x200m2
• Single Access Point on Center
• Codec G.723.1(6.3kbps)
• 24 Payload at every 30ms
• No Silence Suppression Assumed
• Full-duplex Traffic
• RTP( Real Time Protocol) on UDP

19
Simulation Setup(3)

• Random Arrangement

20
Utilizing the Trace files

• NS2 generates a trace file which contains the
  simulated data.
• This data needs to be sieved through using
  applications like XGraph or NAM
• Another way - by making the tools in AWK or C
  which harness the trace files data.
• We have implemented a tool in C to search the
  data generated by NS2 for information regarding
  the throughput and lost data.

21
Statistical Results of the Simulations
22
Simulation Results(Contd.)
23
Comparison between the Simulation and the actual
Measurement
24
Conclusion

• The measurements done using Chariot give us a
  good estimate of the maximum number of VOIP
  sessions that can be supported by one access
  point (using the 802.11 mac layer protocol)
• However due to the probabilistic modeling of the
  call arrival times and the call duration times
  the NS simulations give us a realistic estimate
  of the capacity.
• Thus we conclude that the 802.11b protocol can
  support 14-18 simultaneous VOIP sessions using
  the G.723.1 codec and 8-10 VOIP sessions using
  the G.711 codec.

25
Future Work

• The queue length - packet drop rate trade-off
• We tried various simulations with upto 20 times
  increase in the queue length.
• Shortfalls of the existing 802.11b protocol.
• If the physical layer is implemented using CDMA,
  performance upgrade could be dramatic.