Heterogeneous WiFi Transfer Characteristics Kevin Sinha and Rigoberto Núñez ECE 4605 Final Presentation Thursday, December 7, 2006

Introduction Context: Wireless networks are everywhere and are still increasing in number, it is important to distinguish the quality of available connections Motivation: To see if we are meeting the need for higher throughput rates, stronger security, and better quality of service Problem Statement: How do the throughputs of 802.11 b/g/n wireless technologies measure up in real-world environments?

Background Research Differences between b/g/n 802.11g with MIMO Ideal Range (indoor) ~150 feet ~100 feet ~160 feet Up to 3x 802.11g (300 ft.) Ideal Throughput 11 Mbit/s 54 Mbit/s 540 Mbit/s Up to 8x 802.11g (432 Mbit/s)

Background Research Test Equipment and Environments 2 laptops Linksys router (b/g), Linksys SRX router (MIMO – “n”) Netgear card (b), Linksys card (g), Linksys SRX card (MIMO – “n”) 350 watt DC to AC inverter Software: DU Meter, Bandwidth Monitor Pro, WatchWAN (didn’t work) Environments Outdoor: ex. Wal-mart? Office Depot? Indoor: ex. Interior of parking lot?

Work Performed Background work – collecting WiFi networking data and equipment Original tests: Indoor vs. Outdoor 2-floor Elevator New tests: Vehicle Mobility TCP vs. UDP Times of the day (Centergy Building)

New or Revised Slide Vehicle Mobility Goal: To see how throughput was affected when data was transferred wirelessly between two vehicles in motion 2 cars in the Office Depot parking lot Tested in 4 sets for each of b/g/n – 10 miles/hour following, 25 miles/hour following, 15 miles/hour opposite direction, and 75 miles/hour on I-75 5 loops each parking lot set Parking area photo and diagram

Vehicle Mobility – 10 miles/hour 802.11g 802.11n

Vehicle Mobility – 25 miles/hour 802.11g 802.11n

Vehicle Mobility – Opposite 802.11g 802.11n

Vehicle Mobility – 70 miles/hour New or Revised Slide Vehicle Mobility – 70 miles/hour We decided to go ahead and tackle one of our “future work” goals by taking our vehicle mobility testing to the highways of Atlanta Chose three paths along I-75 that were approximately equal in length (about 4.5 miles) Maintained an average speed of 70 miles/hour Because the tests were performed in an environment where we had minor control, we did them twice each and picked the “better” data to discuss Maps for our paths are shown on the next slide – arrows indicate direction of motion, and green/red dots indicate the points where we started/stopped data collection

Vehicle Mobility – 70 miles/hour New or Revised Slide Vehicle Mobility – 70 miles/hour 802.11n 802.11b, 802.11g

Vehicle Mobility – 70 miles/hour New or Revised Slide Vehicle Mobility – 70 miles/hour 802.11b 802.11g 802.11n

Vehicle Mobility – 70 miles/hour Results Discussion New or Revised Slide Traffic varied from a line-of-site connection between our two vehicles to 2-3 cars in between our vehicles, which resulted in a good representation of the real environment Traffic influence on the data can be seen at the low points, where interference in the signal caused dips and/or drops Interestingly, b and g averages were better at 70 miles/hour than our 10 and 25 miles/hour tests – this is probably due to environment in which our 70 mile/hour test the cars were always “aligned” (one behind the other), whereas in our 10 and 25 tests, the cars had significantly less harmonious motion (sharp turns, driving at angles, etc.). However, the MIMO test averages decreased in performance from b->g->n, which may reflect their design differences (use of multiple antennas)

Vehicle Mobility – Overall New or Revised Slide

New or Revised Slide TCP vs. UDP Goal: To see how throughput differed when transferring a file directly (via TCP) versus viewing it in a streaming format (UDP) UDP tests were performed using VideoLan VLC media player and monitored using Wireshark to ensure UDP connection Indoor tests were redone to ensure absolute similarity in the test environment (only other network in the area was GTwireless at a 30% connection-availability) Maintained constant 25ft distance Used 500mb Video File – Taxi Driver DVD-Rip (Legal)

New or Revised Slide TCP vs. UDP - Results

Times of the Day Centergy Building Goal: To see how concentration of networks and business-day network activity affects throughput Tested in an area of extremely concentrated WiFi network activity (5th floor hallway) 3 times of the day: approximately 10:30AM, 3:30PM, and 7:30PM Monitored all other networks, and performed throughput measurements on both Channel 1 and Channel 6 Maintained 25ft distance for all testing

Times of the Day 10:30am Results

Times of the Day 3:30pm Results

Times of the Day 7:30pm Results

Times of the Day Overall Results

Summary and Conclusions New or Revised Slide Summary and Conclusions Conclusions Many results consistently followed theoretical limitation-estimates, but a couple of interesting anomalies Mobility testing: Vehicle distance has more influence than speed Time of Day testing: N fluctuates noticeably more than B/G, which performed similarly At high-speeds (highway testing), the interference of traffic played a large role in the ability to transmit Lessons Learned Ideal test locations with good line-of-sight, limited traffic, and few network connections are difficult to find Unexpected environmental factors: equipment failing, rough weather (rain) Future Work Multiple users, Additional interference testing

Questions?