A model for traffic arrival in a computer network is analyzed. This model is based on a train arrival process in which the traffic between each node pair consists of a certain number of trains, each having a random number of packets. With this scheme, the number of overhead bits per packet is reduced. The processing time at each node is reduced, because each node needs to make only one routing decision per train and does not need to make a separate routing decision for each packet. The scheme described can cope with the ever increasing amount of data being transferred over computer networks by decreasing processing times while maintaining maximum packet size restrictions. The model accurately describes the dependence suggested by observed correlation between successive arrivals of a train. The aggregate arrival process that results from multiplexing N sources to a multiplexer is characterized in detail. The relation of the proposed model to delay congestion and buffer sizes is described.<<ETX>>
[1]
Ward Whitt,et al.
Characterizing Superposition Arrival Processes in Packet Multiplexers for Voice and Data
,
1986,
IEEE J. Sel. Areas Commun..
[2]
Ward Whitt,et al.
Approximating a Point Process by a Renewal Process, I: Two Basic Methods
,
1982,
Oper. Res..
[3]
P. Kuehn,et al.
Approximate Analysis of General Queuing Networks by Decomposition
,
1979,
IEEE Trans. Commun..
[4]
Raj Jain,et al.
Packet Trains-Measurements and a New Model for Computer Network Traffic
,
1986,
IEEE J. Sel. Areas Commun..
[5]
W. Whitt,et al.
The Queueing Network Analyzer
,
1983,
The Bell System Technical Journal.
[6]
Susan L. Albin,et al.
Approximating a Point Process by a Renewal Process, II: Superposition Arrival Processes to Queues
,
1984,
Oper. Res..