HSLTP—An LTP Variant for High-Speed Links and Memory Constrained Nodes

Abstract

Delay/disruption-tolerant networking architecture relies on the use of Licklider Transmission Protocol (LTP) on interplanetary links. LTP loss recovery is based on automatic repeat request retransmissions, which, when the propagation delay is very long, are costly. Alternatively, losses could be recovered by using Packet Layer Forward Error Correcting codes, as done by the authors in ECLSA (error Coding Link Service Adapter), recently presented in a companion paper, where LTP segment retransmissions are limited to the unlikely case of decoding failures. However, on high bandwidth-delay-product links, the very possibility of segment retransmission requires that a huge number of Rx buffers be available. To resolve this problem, high-speed LTP, presented here, has a more disruptive approach than ECLSA: it enforces almost one-to-one correspondence between LTP blocks and FEC codewords, and never requires LTP segment retransmissions. In the unlikely case of a FEC failure, the partially received block is discarded and its bundles are resent directly by the bundle protocol. The advantages and disadvantages of this approach are explored in the article. On nodes limited in memory, the results are significantly improved.