Experimental Results on IP-layer Enhancement to Capacity of VoIPv6 over IEEE 802.11b WLAN Presented for WiNMee05 April 3, 2005 Youngjune Gwon, James Kempf,

Experimental Results on IP-layer Enhancement to Capacity of

VoIPv6 over IEEE 802.11b WLAN

Presented for WiNMee’05

April 3, 2005

Youngjune Gwon, James Kempf, Raghu Dendukuri, and Ravi Jain

[email protected] USA Labs

© 2005 by DoCoMo Communications Laboratories USA, Inc. Confidential proprietary. All rights reserved. Youngjune Gwon 2

Summary Motivation

– Limited VoIP capacity of wireless LAN– Limited availability of experimental measurements

• Many existing studies based on analytical modeling or simulations

– Unstudied impact of IP or higher layer enhancements• Most focused on modifying 802.11 MAC

Our contributions– One of the first experimental results for VoIPv6 over

802.11b– Proposed use of IP or higher layer enhancement schemes

• ROHC and Silence Suppression (SS)– Analytical capacity model using ROHC, SS, IPv6, 802.11b,

and G.711 parameters– Experimental validation

Key technical results– Achieved best-case capacity improvement of 140 %

without MAC modification or PCF– Analytical model more optimistic about capacity than the

actual by 20%


Outline Motivation Related work Capacity enhancement Hole and Tobagi upper-bound Experimental methodology Results Conclusion


Motivation Today’s WLAN is being used not only for data

(TCP) but also for voice and other real-time multimedia

Maximum capacity of VoIPv4 over WLAN (Vo4WLAN) is 6

– Assuming G.711 with frame duration of 10 msec and 802.11b

Higher layer solutions to improve VoIP capacity unstudied

– Compared to many existing enhancements by modifying 802.11 MAC

Limited availability of experimental measurements

– Few in Vo4WLAN– What about VoIPv6 over WLAN (Vo6WLAN)?


Related Work

Capacity Evaluation Studies

Capacity Enhancement

SchemesAnalytical/simulation-based

Experimental

802.11 MAC modifying*

Higher layer enhancing

Vo4WLAN

Garg and Kappes (ICC’03), Hole and Tobagi (ICC’04)

Anjum et al. (GLOBECOMM’03), Newmann (Network World Fusion)

Aad et al. (INFOCOM’01), Suzuki et al. (PIMRC’01), Veeraraghavan et al. (INFOCOM’01), Yeh et al. (ICC’02), Anjum et al. (GLOBECOMM’03)

Any work?

Vo6WLAN

Any work?

Our work Our work

(*) IEEE 802.11 Task Group E (TGe)formed to define 802.11e


Capacity Enhancement Robust Header Compression (ROHC) – RFC

3095– 60 byte RTP/UDP/IPv6 headers fully compressed to

1 – 2 bytes

Silence Suppression (SS)– More than 50 % of telephone speech is silent– Enabled by Speech Activity Detector (SAD)

• When silence detected, need not transmit packets (perfect SS), or transmit minimal bits only

Access router/AP(ROHC compressor)

Mobile wireless host(ROHC decompressor)

Wireless link

Voice payload + RTP/ UDP/ I P headers (uncompressed)

Voice payload + ROHC Header (compressed) Voice payload + Restored RTP/

UDP/ I P headers (uncompressed)

Reference header

Reference header


Hole and Tobagi Upper-bound

N = upper-bound VoIP capacity in 802.11 R = generated voice packets / sec Tvoice = voice packet transmission time over wireless link SIFS = short interframe space TACK = acknowledgment frame time DIFS = DCF interframe space TSLOT = slot duration CWMIN = number of minimum random slots picked during backoff

22

1

MINSLOTACKVOICE

CWTDIFSTSIFSTR

N

Total voice packet transmission timeover WLAN link from caller/callee

CSMA/CA wireless mediumaccess time


Analytical Vo6WLAN Capacity Hole and Tobagi upper-bound capacity for

Vo6WLAN– Assume VoIPv6 over 802.11b, G.711 with 10 msec

voice frame (80-byte voice payload), ROHC (header = 1.5 byte), perfect SS

Vo6WLAN schemes

1Plain

2ROHC only

3SS only

4ROHC +

SS

Upper-bound capacity, N

6 7 13 14

Capacity improvement

- 17 % 117 % 133 %


Experimental Methodology Real wireless IPv6

testbed Voice traffic

– Speech generation

• ITU-T Recommendation P.59

– G.711 codec– Carried over

RTP/UDP/IPv6 Data traffic

– Based on a web traffic model

• Choi and Limb (ICNP’99)

• Observed peak per client data rate of 30 kbps

– Carried over HTTP/TCP/IPv6

Co-located

Access router

I EEE 802.11bAccess Point

I Pv6 router

…….14 mobile wireless I Pv6

hosts

I Pv6 hosts

I Pv6 hosts I Pv6 hosts

Webserver

ROHCComp/decomp

ROHCComp/decomp


Results

Experimental notes– Started from 5 calls and incrementally add a call to 14 total calls

• Call = caller/callee pair of one wireless and one wired nodes– Number of web clients matches number of wireless nodes– Captured all transmitted and received packets

• Measuring packet loss and latency– Computed MOS for each case

• Used ITU E-model• MOS > 3 to determine capacity

Validation of analytical model– More optimistic than experimental results by 20 %

Capacity improvement– 17 % (ROHC) to 140 % (ROHC + SS)

Background web traffic

Vo6WLAN schemes

1Plain

2ROHC only

3SS only

4ROHC + SS

No 6 (6) 7 (7) 12 (13) 13 (14)

Yes 5 6 10 12Note: values in ( ) represent analytical upper-bound


Obtaining Experimental Capacity

5 6 7 8 9 10 11 12 13 141

1.5

2

2.5

3

3.5

4

4.5Downlink MOS vs. total number of voice sessions

Number of VoIP sessions (calls)

MO

S

No ROHC, no SS, no web traffic (I)

No ROHC, no SS, with web traffic (II)With ROHC, no SS, no web traffic (III)

With ROHC, no SS, with web traffic (IV)

No ROHC, with SS, no web traffic (V)

No ROHC, with SS, with web traffic (VI)With ROHC, with SS, no web traffic (VII)

With ROHC, with SS, with web traffic (VIII)


Conclusions Presented experimental VoIPv6 over 802.11b Proposed use of IP or higher layer

enhancement schemes– ROHC and Silence Suppression (SS)

Instantiated analytical model for VoIP capacity in wireless LAN

– Using ROHC, SS, IPv6, 802.11b, and G.711 parameters

Validated experimentally Key technical results

– Achieved best-case capacity improvement of 140 %

– Vo6WLAN capacity predicted by analytical model is off by 20 %

Documents

Experimental Results on IP-layer Enhancement to Capacity of VoIPv6 over IEEE 802.11b WLAN Presented for WiNMee05 April 3, 2005 Youngjune Gwon, James Kempf,