A project in cooperation with:
Wireless networks deviate from wired ones in many properties. The bit-error rate is a magnitude higher, the medium is shared by many entities, and the data rate is lower, to name a few. The Transmission Control Protocol (TCP) has been designed with fast, low bit-error rate networks in mind. Many of the TCP variants, often called "flavors", adapted TCP to networks with a high bandwidth-delay product. These modifications try to keep the virtual channel filled at all times with the maximum number of segments in flight. Coping with the high delays is the focus of research. In contrast, wireless networks exhibit much higher error rates and often times mobile network entities. Lost segments are usually not a sign of internetwork congestion but due to the properties of the medium. Many approaches exist for this application domain: Mobile TCP, Snooping TCP, I-TCP, Westwood+, and many more. The TCP-Proxy project focuses on a specific application scenario that cannot be solved with mentioned ones.
The radio technology to be used in the application scenario provides data rates between 75 bps and 9600 bps with delays in the scale of 1 s to 20 s. Timeouts are the result when using TCP as congestions is assumed, thus a TCP-Proxy shall be used with every host to provide the following features:
- The TCP-Proxies are transparent to all communicating hosts
- No TCP parameter changes of hosts
- Suppresses TCP timeouts
- Adopts TCP flows to low data rate
- Any kind of higher level protocol can be used
- Preserve TCP end-to-end semantic
As with I-TCP the end-to-end connection is broken into three parts. Each host communicates with a foreign host via a local TCP-Proxy. The following figure shows the network architecture:
One technical report is scheduled for publication in January, 2009.