Wireless Networks

In the next Tuesday meeting (May 30, 1995, 10am, GVU conference room) of the future computer environments group I would like to discuss our needs for mobile/wireless networking. The goal of the discussion will be to make a decision on what systems to buy. I will start the discussion by outlining our options, and then discuss "typical" applications: web browsing and location/orientation aware tour guides.

Mobile Networking Options:

Spread spectrum radio LANS (at 900MHz or 2.4GHz) are available from a number of vendors. The interface to the mobile computer can be through a parallel port or a PCMCIA card, and there is an additional unit (variety of sizes) and antenna that must also be carried around. Within a "cell" maximum data rates of 1-2Mbps (megabits/sec) are possible. Just as with a wired LAN, this data rate is shared by all the users in the cell. The cost per mobile unit is $700-$1000, and the number of users/cell ranges from 15-50. Note that if we had a classroom full of 50 active users we would need a cell with that capacity. The cells are large and penetrate walls. Cells could be established outdoors. Unless we reverse engineered the equipment, the vendor provided software may lock us into particular PC operating systems.

Infrared LANs provide a cheaper and less cumbersome solution that only works indoors. PCMCIA or parallel port interfaces exist, and the additional "stuff" is much smaller. Cells are restricted by walls and potentially doorways, and cell sizes range from 17'x17' to 30'x30'. Maximum data rates are 1Mbps. Cost per mobile user is around $400-$500.

When do we need a wireless network? It would be convenient to never have to plug my handheld or notebook computer into a network. However, wireless connectivity is only really required when one is moving around. I can imagine a hybrid solution in which I plugged in every time I sat down, and a wired network connector or a suitable computer was a part of every chair. The overall network would have to manage transitions between actually being mobile using the wireless network and being stationary plugged in to a wired network. Another solution would be to have two types of wireless networks. One type with large cells would support mobile computing, and another network would have very small cells and support stationary computing (my chair would be a cell, and perhaps have the wireless network transceivers built in).

What protocol should be run on the wireless network? It is important that a "slow" network does not make the mobile unit unusable. Options that have been suggested include:

  1. Use whatever PC operating system that is available and its ethernet interface.
  2. Make the mobile devices X servers (using LINUX), and send X graphics commands across the net. This approach makes the mobile devices just interface devices.
  3. Run applications such as Netscape locally. Netscape over a phone line, for example, has much faster performance when run locally than when run remotely as an X client.
  4. Expect all applications to have front ends and back ends (Mathematica, for example) which define unique network protocols.
  5. Develop toolkits to standardize different types of high level interactions.
  6. Make the mobile unit a "Hot Java" virtual machine.
  7. Use Appletalk.

What bandwidth is required? Asymmetric bandwidths are acceptable: a large bandwidth is desirable for moving images to the mobile unit, and a small bandwidth is acceptable in moving human input (selections, pen strokes, input text) out to servers on the net. A guess is that 30kbps to the mobile unit and 1kbps to the net are minimum bandwidths (the 30kbps figure comes from my use of Netscape over a phone line with 28.8k baud (bps) modems). For applications like the tour guide we could install CDROM drives on the mobile units and use write once CDROMs (GVU will have a writable drive) for each new "tour", and network traffic could be used to help select segments on the CDROM, to provide more up to date scheduling and demo information, and to negotiate for reservations and tickets. In this case a 4800baud (bps) link to the mobile unit will probably be acceptable. This rate can be achieved by packet radio networks such as ARDIS, allowing us to implement a tour guide that works outdoors and throughout Atlanta (and also throughout the US) today. A 64kbps link would allow phone quality audio and ISDN compatible networking. A 1-2Mbps link would allow compressed video streams to be sent. A longer range goal is to include a digital still camera and/or video camera with a mobile unit.

Another issue is localization. We would like to have the ability to make mobile units location and orientation aware (they know where they are at some resolution and which way they are pointed). If we could use the mobile network to help us solve this problem (make available information on which cell a unit is in) that would be helpful. Alternatives are attaching PCMCIA GPS and compasses to each mobile unit, or building a separate locator network. Issues:

Click here for a file listing projects, vendors, and products.

Some Readings On Pda Based Web Browsers:

W4 - the Wireless World Wide Web
Joel F. Bartlett - Digital Equipment Corporation Western Research Lab.
Dynamic Documents: Mobile Wireless Access to the WWW
Mobisaic: An Information System for a Mobile Wireless Computing Environment
Geoffrey Voelker and Brian Bershad - University of Washington.
These papers are from a conference: Workshop on Mobile Computing Systems and Applications with many other interesting papers. Other relevant papers on the web:
PDAs as mobile WWW browsers
Stefan Gessler and Andreas Kotulla, Telecooperation Office, University of Karlsruhe
DATAMAN project: Towards a Mosaic-like Location-Dependant Information Service for Mobile Clients
Arup Acharya, Tomasz Imielinski and B. R. Badrinath