The RAPID transport protocol

Abstract

Future networking environments are expected to provide more bandwidth and be more robust to bit errors. Such changes will pose new requirements for data transfer protocols. Rather than trying to adapt an existing transport protocol to this new environment, we decided to develop a new one called RAPID. The most important features of RAPID are its unidirectional data transfer capability and a clear division between signaling and data transfer functions. The paper covers a description of such a new networking environment with a closer look to its error characteristic, which are expected to change from bit errors to packet losses at buffering points. It describes the protocol architecture, presenting the clear distinction made between connections signalling and data transfer functions, as it might be a characteristic of future data transfer systems, also the forward error correction code used by RAPID to increase a connection reliability degree is described. Finally we make a performance comparison between RAPID and a traditional protocol, which enhances a connections reliability degree by making use of retransmissions. For this comparison we use a top down model for the data transfer system based on queues, with particular attention to queues overflow. The results of the model analysis shows that if certain assumptions made about the errors characteristic and time parameters are valid, a unidirectional data transfer protocol is able to provide a higher throughput, by user's means, than traditional protocols. >