Delay/overhead measurements for circuit-emulation tunnels

Abstract

Due to unique security and communication channel constraints, military communication networks often must send data between serial devices at extremely low data rates. As military networks begin to migrate towards packet-based transport, data from these serial applications will need to be packetized and transferred from sender to receiver in a circuit-emulation tunnel. In addition to packetizing and serializing the data sent across the network, these circuit-emulation tunnels must also remove the jitter caused by packet-based transport. Packetizing/serializing data and removing jitter can result in a significant amount of bandwidth overhead and delay. This paper provides a brief description of various circuit-emulation technologies and a detailed analysis of the bandwidth overhead and delay due to serial-to-packet conversion. An analytical model was developed that predicts bandwidth overhead and delays resulting from the circuit-emulation tunnel as a function of the serial interface speed, packet size, jitter buffering, and control messaging. The analytical model was tested against a series of network emulation experiments using commercial circuit-to-packet devices. Results show a fundamental trade between bandwidth overhead and delay. Circuit-emulation tunnels with small packet sizes have short packetization delays and large bandwidth overhead, while tunnels with large packet sizes have long packetization delays and smaller bandwidth overhead. These trends are more pronounced at the lower data rates because it takes a longer time for the serial data to be packetized. Experimental and theoretical results presented in this paper show similar trends with slight discrepancies due to internal processing effects not included in the analytical model.