The Impact of Datagram Partition and Checksuming on Datagram Latency

Abstract

In this paper we present a technique for datagram division into several parts, with one header per datagram, and checksum for every part. The proposed technique is evaluated in terms of improvements in datagram error rate and datagram latency. We show that, with small datagram length enlargement, datagram error rate and datagram latency can be significantly reduced (i.e. with one per cent control overhead we can reduce, approximately 5 times, latency and error rate of datagram parts in 1000 bytes long datagram )