Chen J. Minist Sci & Technol Peoples Republ China Informat Ctr Beijing 100862 Peoples R China. Liu L. X.; Hu X. H. Chinese Acad Sci Inst Software Beijing 100190 Peoples R China.
Abstract:
Incorporating network coding into TCP has the advantage of masking packet losses from the congestion control algorithm. It could make a lossy channel appear as a lossless channel for TCP, therefore the transport protocol can only focus on handling congestion. However, most schemes do not consider the decoding delay, thus are not suitable to be implemented in practical systems. We propose a novel feedback based network coding (FNC) retransmission scheme which has high throughput and quite low decoding delay without sacrificing throughput. It uses the implicit information of the seen scheme to acquire the exact number of packets the receiver needs for decoding all packets based on feedback. We also change the encoding rules of retransmission, so as to decode part of packets in advance. The scheme can work well on handling not only random losses but also bursty losses. Our scheme also keeps the end-to-end philosophy of TCP that the coding operations are only performed at the end hosts. Thus it is easier to be implemented in practical systems. Simulation results show that our scheme significantly outperforms the previous coding approach in reducing decoding delay, and obtains the throughput which is close to the scenarios where there is zero error loss. It is particularly useful for streaming applications.
English Abstract:
Incorporating network coding into TCP has the advantage of masking packet losses from the congestion control algorithm. It could make a lossy channel appear as a lossless channel for TCP, therefore the transport protocol can only focus on handling congestion. However, most schemes do not consider the decoding delay, thus are not suitable to be implemented in practical systems. We propose a novel feedback based network coding (FNC) retransmission scheme which has high throughput and quite low decoding delay without sacrificing throughput. It uses the implicit information of the seen scheme to acquire the exact number of packets the receiver needs for decoding all packets based on feedback. We also change the encoding rules of retransmission, so as to decode part of packets in advance. The scheme can work well on handling not only random losses but also bursty losses. Our scheme also keeps the end-to-end philosophy of TCP that the coding operations are only performed at the end hosts. Thus it is easier to be implemented in practical systems. Simulation results show that our scheme significantly outperforms the previous coding approach in reducing decoding delay, and obtains the throughput which is close to the scenarios where there is zero error loss. It is particularly useful for streaming applications.
Chen J.,Liu L. X.,Hu X. H.,et al. effective retransmission in network coding for tcp[J]. INTERNATIONAL JOURNAL OF COMPUTERS COMMUNICATIONS & CONTROL,2011-01-01,6(1):53-62.
