Satellite networks have been deployed with success by several companies in order to allow global broadband access to highly mobile nodes. It is well known that the standard transport protocol, TCP, is not well suited for data transport over satellite links that are characterized by a relatively high bit error rate (BER) and a long propagation delay. Links characterized by high BER suffers of dramatic drops in term of goodput because of the congestion control algorithm employed by TCP Reno which blindly halves the congestion window (cwnd) when 3 duplicate acknowledgment are received inferring a congestion episode. This heuristic leds to undue cwnd reductions in the presence of lossy links where the duplicate acknowledgment are likely to be due to corruption of data packets.
We will develop and test a modified TCP congestion control scheme based on TCP Westwood+ and customized for satellite networks. The novelty of our approach is that we adapt the cwnd to the current end-to-end estimated bandwidth instead of blindly halving it. It has been reported in previous literature that TCP Westwood+
Satellite networks have been deployed with success by several companies in order to allow global broadband access to highly mobile nodes. It is well known that the standard transport protocol, TCP, is not well suited for data transport over satellite links that are characterized by a relatively high bit error rate (BER) and a long propagation delay. Links characterized by high BER suffers of dramatic drops in term of goodput because of the congestion control algorithm employed by TCP Reno which blindly halves the congestion window (cwnd) when 3 duplicate acknowledgment are received inferring a congestion episode. This heuristic leds to undue cwnd reductions in the presence of lossy links where the duplicate acknowledgment are likely to be due to corruption of data packets.
We will develop and test a modified TCP congestion control scheme based on TCP Westwood+ and customized for satellite networks. The novelty of our approach is that we adapt the cwnd to the current end-to-end estimated bandwidth instead of blindly halving it. It has been reported in previous literature that TCP Westwood+