University of Minnesota Supercomputing Institute Report No. 2001/135
Minneapolis, MN 55415-1227
Wireless Internet Access Across Washington Avenue
Also available as a PDF at http://www.msi.umn.edu/~boggsj/wireless.pdf
John M. Boggs 1 and David A. Yuen2
Minnesota Supercomputing Institute (M.S.I.)
1200 Washington Avenue South
Department of Geology and Geophysics
University of Minnesota, Minneapolis, MN 55415-1227
We have studied the performance of wireless Internet traffic across Washington Avenue in Minneapolis using the Wi-Fi or 802.11b standard. We have compared both the downloading and uploading rates under different environmental situations ranging from rush hour to late at night. We have also found the importance of the increased boosting transfer rate provided by a portable antenna under indoor conditions. Our results show that uploading in general is faster than downloading and that the use of the antenna with a handheld PC can increase considerably the indoor range. The access range of wireless Internet is doubled outdoors. We find that the height of the access point can help to spread the wireless signal enormously. The speed of reception is improved by the use of a faster CPU and bus. Thus wireless Internet traffic on a PC is much faster than that on a handheld PC. This communication should help the university populace understand the current capabilities and limitations of wireless technology.
November 20, 2001
In the last few years there has been a growing rage about wireless technology and its impact on the wired Internet. As Albert Einstein once said:
“You see, wire telegraph is a kind of a very very long cat.
You pull his tail in New York and his head is meowing in Los Angeles.”
Well, wireless Internet operates exactly in the same way; the only difference is that there is no cat on the wireless Internet.
Figure 1
Washington Avenue in Minneapolis. Minnesota Supercomputer Institute is on the left, Maxwell’s Café on the right, looking toward I-35W.
Over the past few months, we (Yuen's Group) have had ongoing experiments with wireless Internet access at MSI and particularly at Maxwell's Cafe across Washington Ave. This was made possible through the use of the 802.11b wireless networking protocol and small handheld computers. Basically, 802.11b (or Wi-Fi) is the same as a standard wired network but without the wires. Its range can vary greatly depending on the antenna and transmitter setup, from around 300 meters to several tens of kilometers. With the current version the theoretical maximum speed is 11 megabits per second.
Figure 2
Wireless client connects to the nearest available wireless access point.
Many people confuse this technology with cellular phones. While 802.11b wireless networking operates at a 2.4GHz radio frequency, the same as most cell phones, it has nothing to do with the cellular network. In order for it to work, it requires a "base station" to be installed at a location that is close to where you want to get access. This base station needs to be plugged into a regular wired network. It then relays data between whatever wireless devices are talking to it and the rest of the network (see Figure 2). The rest of this report will be concerned with the technical details of our setup, such as software and hardware configuration, some experiments we performed and problems we ran into. We hope that this holistic approach will help many of the novice users to understand their own needs.
Technical Details
When we first decided to go wireless in July, 2001 we started with a Linksys WAP11 access point and a few Avaya Gold wireless PCMCIA cards used with Hewlett-Packard Jornada 720s. We also have an Apple Macintosh G4 system configured as an Airport base station (Airport is just Apple’s name for 802.11b wireless) but it is only useful inside MSI so will not be discussed in detail. We later acquired an Intel 2011 wireless access point and added that to our network. We found that generally the Intel access point performed better than the Linksys access point and that will be discussed in more detail below.
Figure 3
Our wireless Jornada Setup with the wireless card outside (left) and connected (right) to the Jornada 720.
On the left are the three separate components: Wireless PCMCIA card, Jornada, and the antenna from Lucent technology. On the right are the three parts together. The wireless card fits almost completely inside the Jornada. The antenna is optional, but can make a big difference in marginal situations.
A Jornada 720 from Hewlett-Packard (HP) is a small handheld touchscreen computer that runs the Windows CE operating system. It has most of the capabilities of a full size laptop, but folds up to less than the size of a VHS videotape, and has an 8-hour battery life. HP makes another product called Jornada, but that is the 500 series and is a Palm Pilot-like device with no keyboard but can run much of the same software as the Jornada 720. It has a 206Mhz StrongArm RISC processor and 32MB of internal RAM. The RAM is split between storage space and program memory and this is adjustable through software. However, storage can be almost limitless because the Jornada has a Type II PCMCIA slot and Type I CompactFlash slot. The CompactFlash slot can only be used for storage expansion, and CompactFlash cards are available in capacities of up to one gigabyte. The PCMCIA slot can also be used for storage, but also for other things such as wireless networking cards, standard (wired) network cards, modems (although the Jornada has a built-in modem), GPS cards, and much more. We also did testing with a 700Mhz Dell laptop with 256MB RAM and a 500Mhz Apple PowerBook G4 laptop computer with 1024MB RAM. Encryption on wireless is currently set at 64 bits and 128 bits depending on the price of the PCMCIA cards. Flaws in the encryption have been noted (Borisov, et al. 2001), but the working environment is the overall consideration, since one can be in a library, a park or a café. We note that ssh is far more secure than the encryption currently provided on the PCMCIA wireless cards and on the base stations.
Figure 4
The Linksys WAP11 Access point on the left and Intel 2011 Access Point on the right.
The two wireless access points are used to provide access to the wired network. No special network configuration is required other than an IP address so the access point can be remotely administered. The access point is transparent to the network; it merely routes network traffic between wireless clients (such as our Jornadas) and rest of the network. The wireless clients are assigned IP addresses by the University of Minnesota’s DHCP service ( https://wwws.nts.umn.edu/cgi-bin/dhcpreg?opt=aup ), so no special configuration is required for the clients either, except for installing the driver for the PCMCIA card, which is trivial. The wireless card must be registered for use with the DHCP service at the University of Minnesota.
The software for the Avaya wireless card provides some feedback as to how strong the radio signal is and therefore how fast your connection is. There is a signal to noise ratio meter (SNR) and 3 keywords associated with the signal to noise ratio SNR: Marginal, Good, and Excellent. We found that for interactive software such as ssh or ftp a Good or Excellent signal is needed to do any useful work. Marginal is acceptable for non-interactive software such as a web browser or email program. These descriptors for the signal strength will be quantified below. The Avaya wireless cards can be used with any laptop or Handheld PC and come with software for many Windows CE devices.
We later added Avaya wireless Range Extender Antennas that plug into the PCMCIA cards to improve the wireless reception and transmission. See Figure 3 for a photo of our setup.
Configuration and Experiments
There are very many factors that can affect the performance and operation of a wireless network. These include location of the access point, distance, and structures blocking the signal such as walls and people, time of day, traffic on Washington Avenue, temperature, and software issues. Distance is the obvious limiter; the farther away from the access point you get, the weaker (and slower) your signal is. Blockage from walls and moving vehicles will weaken the signal, especially if there is metal in the wall or metal structures around the wireless client. People getting in the way of the signal will also degrade it or block it entirely. This is where the range extender antenna comes into play. It can often overcome these obstacles and get a usable signal. The wireless signal does not seem to be affected by glass windows.
Figure 5
Wireless Internet at Maxwell’s Café. Notice the screen showing cnn.com. This is a live shot.
Our first experiment with wireless at Maxwell’s was in July. We put the Linksys access point in Dave Yuen’s office in the window. We were only able to get a marginal signal on the sidewalk outside of Maxwell’s as most of the signal was blocked by the building structure of MSI. Since the window in Dr. Yuen’s office does not face Maxwell’s, there is no direct line-of-sight across Washington Avenue. Also the signal would never last more than one hour before the access point overheated because of the greenhouse effect caused by it being between the window and blinds. It became very hot in that window when the sun was out. Resetting the access point would get it going again, but it would always lock up after a short time. It worked much better at night after sunset; then it stopped locking up. Traffic could also affect the signal both during the day and at night. When there was a lot of traffic, especially with big trucks, it was difficult to establish a signal. Presumably the radio waves were scattered by the moving vehicles and not received.
Later we moved the Linksys access point into Ann John’s room which is at the front of MSI and her window has a direct line-of-sight to Maxwell’s. The firmware (internal software) of the access point was upgraded to the latest version, and we put the access point in her window. Reception at Maxwell’s was much better, and actually worked well inside Maxwell’s. However, we were still having the same overheating problem as before. We removed the access point from the window and put it on a shelf facing the window. This seemed to help, but it was still locking up occasionally. We put the original version of the firmware back on instead of the new one, and this solved the lockup problem. Since then it has been working fine.
We put the Intel access point in Kay Anderson’s office by the window. Kay’s office is right next to Ann’s, so it also has line-of-sight to Maxwell’s. The Intel access point has many more features than the Linksys such as operation in repeater mode and remote administration through telnet, but these extra features are not relevant to our experiments. However, all those extra features make the Intel access point more difficult to configure, and the documentation that came with it is poor. But once it was working, it worked better than the Linksys; it is able to penetrate farther back into Maxwell’s Café (it even works from the bathroom all the way in the back if you use the range extender antenna!). The Intel access point poses no problem for many Internet users at Maxwell’s. The outdoor range is extended considerably by the antenna. It is approximately doubled to around 100 meters from Ann Johns’ window. This access point also seems to be immune to overheating.
Figure 6
Diagram of wireless networking across Washington Avenue. It is about 60 meters from the location of the access points and the front window of Maxwell’s Café. The length of Maxwell’s seating area is about 20 meters.
The main test used to compare the two access points at Maxwell’s was the data rate of transferring files over ftp. As you can see from Figures 7a and 7b, the Intel access point is faster. Two of the data sets were taken with Jornadas and the third was taken with a Dell laptop PC. The third data set is incomplete because we were able to get a signal from only the Intel access point using the range extender antenna because heavy traffic on Washington Avenue caused too much interference. Transmission rate on the Dell PC is faster, probably because of a faster processor and bus speed, or more memory for a buffer during file transfer. We tried to get a signal with an Apple PowerBook G4 but were unable to obtain any signal with its built-in Wi-Fi card, and there is no place to plug in an external antenna.
Figure 7a.
Graph of transfer rates for uploading from Jornada to MSI. ftp gives transfer rates in Kbytes/sec. The size of the file is given in terms of Mbytes. For example, 2.1 above the vertical bar means the test file was 2.1 Mbytes in size.
Figure 7b.
Graph of transfer rates for downloading from MSI to the Jornada. Ftp gives transfer rates in Kbytes/sec so that’s what is used here. The size of the file is given in terms of Mbytes. For example, 2.1 above the vertical bar means the test file was 2.1 Mbytes in size.
Network technology has been advancing very quickly the past two or three years. The first wireless networks were slow, less than 1 megabit/second. Now we have 11 megabits/sec. Soon a new standard will be available for wireless, 54 megabit/sec. If networking technology keeps advancing at this rate, networks will need to be upgraded every few years. This could become expensive.
In this report we have shared our experience accumulated in the last few months of working with a wireless Internet connection, which reaches across the 50-meter stretch of Washington Avenue. Most of our work at this “wireless Internet café” has been based on using a handheld PC, the Jornada 720. First, the location of the access point is very important for providing a wide coverage. By placing the access points at a height of around 15 meters, we were able to reach clear across Washington Avenue. Second, we found the functionality of a keyboard to be essential in monitoring our runs on the supercomputer computers. Third, the growing availability of software, such as LINUX on a Palm Pilot and BSD UNIX on a Jornada, will increase the usefulness of wireless connections to large computing servers. From a programming point of view, we note that network protocols such as ssh (security shell for logging on to the main computer) and ftp (for transferring files) are already available on a Jornada and there are image viewers for looking at jpeg and postscript files. However, we note that there is a trade-off between the connection speed and the range of the access point. The effective range decreases with faster connecting speed. Thus there will be more need for a more powerful antenna with the next generation of 54 Mbps wireless transfer rates.
Because today the Wi-Fi technology is still fairly new, except for metropolitan areas such as New York and Chicago, it can be sometimes difficult to locate a wireless network. At the University of Minnesota you would go to http://lighthouse.micro.umn.edu/kiosk/location.asp to find the wireless access points. Still there are not many and in many departments none at all. In general you should consult http://www.wifinder.com to find the available networks in your neighborhood. But these are not always up-to-date. In New York City, there is a project aimed to provide free wireless access to the public (http://www.nycwireless.net). Undoubtedly there will be in the near future a growing market for a denser coverage of wireless networks everywhere because of its impact on all walks of life.
Wi-Fi: Short for Wireless Fidelity. This is another name for the 11 megabit/sec 802.11b wireless standard. It operates at 2.4 GHz and works by transmitting data rapidly over radio waves.
Encryption: Currently there are two forms of encryption, 64bit and 128bit called Wired Equivalent Privacy (WEP). This encryption is weak and easily broken. It also slows down transfer speed because the data must be encrypted and decrypted. We have not used encryption, and neither does the U of MN. Wireless networks should be assumed to be open, and software like ssh should be used if working with sensitive data is necessary.
Handheld PC: HPC or H/PC for short. A small handheld computer in a “clamshell” design, where the screen folds down over the keyboard. An HPC runs Microsoft Windows CE, which can be activated instantly, without waiting for a booting process. Our Jornada 720s are Handheld PCs.
Pocket PC: Another kind of handheld Windows CE computer but without intrinsic keyboard capabilities. This is shaped like a Palm Pilot, with no keyboard and a smaller screen.
Windows CE: Best described as a “stripped-down” version of Microsoft Windows. WinCE for short. (Microsoft officially denies that “CE” stands for anything.) The kernel is based on the Windows NT kernel. Windows CE boots in a matter of seconds, versus several minutes for any other version of Windows. When an HPC or Pocket PC is turned off, it is not really off, but “asleep” in a low-power mode, so when it is turned back on, it comes up in less than a second. The only way to truly turn off most Windows CE computers is to remove the battery.
Infrared: File transfer using infrared radiation. Has a range of about 1 meter and 115 Kbytes/sec speed.
Ethernet / Wired Network: Most networks are wired with standard Ethernet. The speeds can be 1000 Mbits/sec, 100 Mbits/sec, or 10 Mbits/sec. The Intel and Linksys access points connect to the wired network at 10 Mbits/sec.
PDA (Personal Digital Assistant): PDA is a generic term used to refer to devices like the Jornada 720. PDAs are generally used to store phone numbers, appointments, etc. Some examples are the PalmOS devices made by Palm, Handspring, Sony and others. More powerful PDAs run Windows CE such as our Jornada 720s. Windows CE PDAs that are in the Palm form factor (i.e., smaller screen, no keyboard) are called Pocket PCs. Examples include the Compaq iPaq, Casio E115 and HP Jornada 500 series.
Client – Server: A client is a computer that receives results processed on a server. A server processes the data and makes it available to clients. A web server is an example of a server, and someone using a web browser to connect to that server is a client. Generally servers are more powerful than clients, but this is not always true. For example, it is possible to run the Apache web server on a Jornada 720.
Transfer Rates: Currently Wi-Fi runs at 11 Mbits/sec. For comparison, most Ethernet networks run at 100 Mbits/sec, most cable modems run at 1 or 1.5 Mbits/sec (or less) and 56k modems run at 56 Kilobits/sec, or about 0.056 Mbits/sec. Infrared transfer is 115 Kbits/sec, or 0.115 Mbits/sec. We note that transfer rates are measured in bits, while file sizes are measured in bytes. There are eight bits in one byte, so divide the transfer rates, quoted in Mbits/sec by 8 to get the rate in Mbytes/sec.
Bergeron, Bryan P. The Wireless Web: How to Develop and Execute a Winning Wireless Strategy, McGraw-Hill (2001).
Borisov, Nikita et al. (In)Security of the WEP Algorithm. http://www.isaac.cs.berkeley.edu/isaac/wep-faq.html , (2001).
Garbow, Zachary A. et al. Interactive Web-Based Map: Applications to Large Data Sets in the Geosciences, Electronic Geosciences 6, 2001 http://link.springer-ny.com/link/service/journals/10069/free/technic/webbased/index.htm
Hassibi, Babak. Space-Time Processing for Wireless Communications. http://www.ima.umn.edu/talks/workshops/8-8-10.2001/hassibi/hassibi.pdf , (2001).
Kumar, P. R. Wireless Networks: Analysis, Protocols and Architecture. http://www.ima.umn.edu/talks/workshops/8-8-10.2001/kumar/kumar.pdf , (2001).
McPherson, Frank. How to Do Everything with Your Pocket PC and Handheld PC, Osborne/McGraw-Hill (2000).
Tse, David. Multiuser Diversity in Wireless Networks: Smart Scheduling, Dumb Antennas and Epidemic Communication. http://www.ima.umn.edu/talks/workshops/8-8-10.2001/tse/tse.pdf , (2001).
We thank Fabien Dubuffet and Erik O. Sevre for their help in timing and their interesting ideas. We take this opportunity to thank Ann Johns of DTC and Kay Anderson of UMSI for allowing us to use their offices as part of this experiment. Thanks are also due to Emma Rainey, Zack Garbow, Gordon Erlebacher and Lilli Yang for their encouragement and friendly advice. Tomo K.B. Yanagawa is to be credited with the observation of the influence of car traffic on the wireless reception. We are also appreciative of the splendid help at Maxwell's from Kristen, Courtney, Kari, Angie, Billy and Mary. This research has been supported by both the National Science Foundation and the Department of Energy.
The web counter below was obtained from www.digits.com