Research issues for transport protocols in satellite IP networks

problems of transport protocols in satellite IP networks along with the most recent solutions proposed in the literature. Non-real-time applications use Transmission Control Protocol (TCP). Current TCP schemes have throughput problems in networks with long propagation delays and relatively high link error rates such as satellite networks [1]. TCP throughput decreases mostly because: • The long propagation delays cause longer duration of the slow start phase during which the TCP sender may not use the available bandwidth. • TCP was initially designed to work in networks with low link error rates (i.e., most packet losses were due to network congestion). As a result, the TCP sender decreases its transmission rate. However, this causes unnecessary throughput degradation if packet losses occur due to link errors. In a shared network such as the Internet, real-time traffic flows are also expected to be TCP-friendly [2]; that is, they should comply with the following rules: • Rule 1: Their transmission rate can increase slowly as long as the network is not congested. • Rule 2: Their transmission rate must decrease immediately when the network is congested. Next-generation IP routers will penalize traffic flows not compliant with these rules. For real-time applications there have been several studies in the recent past [2–7]. The transmission rate value is adjusted in such a way that rules 1 and 2 are satisfied. In order to comply with rule 2, traffic sources decrease their transmission rate when packet losses are detected because these are the only congestion signal in the current Internet. However, in satellite networks packet losses can occur due to link errors with probability even higher than 10 –2. If the source decreases its transmission rate when a packet loss occurs due to link errors, the network efficiency decreases drastically. In the beginning of a new connection, TCP has no information about the traffic load on the connection path; thus, it executes the slow start [8] to probe the availability of band-width along the path. The congestion window, cwnd, which gives the maximum number of unacknowledged packets 1 the sender can have in transit to the receiver, is set to one and then incremented by one for each received acknowledgment (ACK). For TCP connections the transmission rate, b, is approximately b ª cwnd/RTT, (1) where RTT is the round-trip time. Therefore, the time required by slow start to reach a bit rate B is t SS ª …