Implementation of Test Bed for Dynamic Channel Selection

Implementation of Test Bed for Dynamic Channel Selection In WLANs Communications Laboratory TKK/HUT

WLANs – Increasing Popularity l Growing Popularity of WLANs l Inexpensive and Flexible l Growing Trend in ad hoc networking l Easy to configure

WLANs – Shortfalls and Issues l Intrinsic unreliable nature of the wireless channel l Unreliable and unpredicable Transmission medium l Speeds less than wired networks l Security

WLAN Standards l l l IEEE 802. 11 - 1 Mbit/s and 2 Mbit/s, 2. 4 GHz RF and IR standard (1997) IEEE 802. 11 a - 54 Mbit/s, 5 GHz standard (1999, shipping products in 2001) IEEE 802. 11 b - Enhancements to 802. 11 to support 5. 5 and 11 Mbit/s (1999) IEEE 802. 11 g - 54 Mbit/s, 2. 4 GHz standard (backwards compatible with b) (2003) IEEE 802. 11 h - Spectrum Managed 802. 11 a (5 GHz) for European compatibility (2004)

Contribution of Thesis l Comparative Study of DFS in 802. 11 b Vs Static Channels In Ad hoc Networks ¡ Multiple radio interferences on a limited Bandwidth ¡ Multilple networks on a Single Channel ¡ Degradation in Throughput ¡ Inflexibility of Channel allocation ¡ Better quality link possible of unused channel

The Solution l Creating a Dynamic Channel Selection (DSC) Mechanism for WLANs in the 2. 4 GHz band l Providing a testbed to obsreve the Improvements offered by the use of a DSC Application l Analyse the improvement in Throughput

Dynamic Frequency Selection in WLANs l Provided by the IEEE 802. 11 h extention to the IEEE 802. 11 a standard l No mechanism currently being employed in IEEE 802. 11 b/g WLANs

A Simple DFS Algorithm

Channel Deployment Issues in the 2. 4 GHz band l. A total of 11 channels in both IEEE 802. 11 b and IEEE 802. 11 g standards l Limited to 3 usable channels due to the interchannel interfernces l Limits the DSC scheme to effectively only switch between 3 channels

Channelization scheme for IEEE 802. 11 b

Setting Up Test Enviornment l Pentium III Desktop PCs with Realtek 802. 11 b/g wireless lan cards l Ubuntu v 5. 10 linux l wireless_tools. 28 toolkit from IBM l Traffic Generator IPerf l Shell Scripting Knowledge l Patience to install WLAN drivers on linux l Configuring WLANs

Lab enviornment l l l l l l l l l l iwlist wlan 0 scan wlan 0 Scan completed : Cell 01 - Address: 00: 14: BF: E 6: 53: 5 E ESSID: "dtn_demo" Mode: Master Frequency: 2. 412 GHz (Channel 1) Quality=37/100 Signal level=12/100 Noise level=0/100 Encryption key: off Bit Rates: 54 Mb/s Cell 02 - Address: 00: 16: B 6: 5 B: E 4: A 4 ESSID: "aalto" Mode: Master Frequency: 2. 412 GHz (Channel 1) Quality=38/100 Signal level=13/100 Noise level=0/100 Encryption key: off Bit Rates: 54 Mb/s Cell 03 - Address: 00: 16: B 6: 5 B: CB: FB ESSID: "aalto" Mode: Master Frequency: 2. 437 GHz (Channel 6) Quality=32/100 Signal level=5/100 Noise level=0/100 Encryption key: off Bit Rates: 54 Mb/s Cell 04 - Address: 42: DC: B 9: 77: 91: 7 B ESSID: "wrt 54 gs" Mode: Ad-Hoc Frequency: 2. 437 GHz (Channel 6) Quality=36/100 Signal level=11/100 Noise level=0/100 Encryption key: off Bit Rates: 11 Mb/s Cell 05 - Address: 6 E: FF: 7 B: 87: 23: 1 B ESSID: "adhoc_test" Mode: Ad-Hoc Frequency: 2. 412 GHz (Channel 1) Quality=62/100 Signal level=47/100 Noise level=0/100 Encryption key: off Bit Rates: 22 Mb/s

Node-A Start DSC Application Channel Quality Monitoring Quality < Threshold No Yes l Text based signalling l Client Server l 3 -way Acks l Link Quality analysis and selection Algorithm l Application Layer Implementation Node-A Channel Change Procedure Send Channel No. to Peer Node Self Channel Change Procedure Delay Probe Channel for Change Request Yes Change Channel& Send ACK Node-B ACK Received No ACK within a Time Frame Send Confirmation Node-A Channel Quality Measurements Start Again/Stop No Receive d Wait for Confirmation

Sequence Diagram for the DCS tool

Test Case – 1 l Comparison of channel performance in a bad channel versus a channel selected by using the Channel Selection Utility for TCP traffic l Intervals of 300, 900, 3600, 7200, 10800, 21600

Data sheet – 1. 1 TCP Traffic Without Channel Selection Utility Time (se c) Data Transferred (Kbytes) 240 42086. 4 1. 43 300. 5 52428. 8 1. 43 600. 5 104448 1. 43 900. 5 157696 1. 43 3600. 5 624640 1. 42 7200. 5 1139160 0. 158 10800. 7 1866465 1. 42 21600. 5 3718705 1. 42 Throughput (Mbits/sec)

Data Sheet – 1. 2 TCP Traffic With Channel Selection Utility Time (se c) Data Transferred (Kbytes) 240. 5 18124. 8 0. 617 300. 4 30617. 6 0. 833 600. 4 82227. 2 1. 12 900. 5 142336 1. 3 3600. 6 615424 1. 4 7717 1290824 0. 167 10800. 8 1939865. 6 1. 47 21600. 4 3845923 1. 46 Throughput (Mbits/sec)

Results l Time consumed in the execution of DCS degrades throughput at smaller intervals due to the silent period l Improvement only seen in times greater than one hour

Test Case – 2 l Comparison of channel performance in a bad channel versus a channel selected by using the Channel Selection Utility for UDP traffic with a continuous data transfer. l Intervals 21600 of 300, 900, 3600, 7200, 10800,

Data Sheet – 2. 1 UDP Traffic Without Channel Selection Utility Time (sec) Data Transferred (Kbytes) Throughput (Kbits/sec) Jitter (msec) 240 30 1. 05 1. 867 1/21401 0. 0046% 300 37 1. 05 0. 297 0/26751 0% 600 75 1. 05 1. 117 0/53501 0% 900 113 1. 05 0. 04 0/80251 0% lost/total %lost UDP Traffic With Channel Selection Utility Time (sec) Data Transferred (Kbytes) Throughput (Kbits/sec) Jitter (msec) lost/total %lost 124 6. 98 473 0. 554 6219/11198 56% 240 14. 3 623 0. 069 8689/21401 41% 300 25. 7 719 0. 261 8403/26751 31% 600 63. 2 884 0. 014 8398/53501 16%

Results l No retrials so lots of lost packets. l Requires a buffer mechanism to be effective to cater when the silent period occurs.

Test Case – 3 l The purpose of this test case is to compare throughput of the radio interface when burst of traffic is generated instead of continuous traffic. l 5 Mbytes of traffic every 5 minutes from 0800 hrs to 1800 hrs l DCS mechanism initated every 20 minutes l Alternatively quality threshold can be used to initate the DCS Mechanism

Throughput over the time interval of 0600 hrs to 1800 hrs, where 5 Mbytes of data is transferred every 5 minutes on the worst channel. Average Throughput 1. 408 Mbits/sec

Throughput over the time interval of 0600 hrs to 1800 hrs, where 5 Mbytes of data is transferred every 5 minutes while the Channel Selection Utility is used 2 -3 times per hour. Average Throughput 1. 467 Mbits/sec

Comparison of the throughputs when the worst channel is in use versus when the Channel Selection Utility is used to select the best channel.

Conclusions l l l l Very distinct Improvement in throughput Implementation on application layer is not efficient Taking advantage of the Draft IEEE 802. 11 k standard for development of DCS mechanim Buffer for UDP traffic during silent period Development of selection algorithms Compatibilty of WLAN drivers in Linux distribution. www. linux-wlan. org Simplification of network configuration needed