Traditionally, models of packet arrival in communication networks have assumed either Poisson or compound Poisson arrival patterns. A study of a token ring local area network (LAN) at MIT [5] found that packet arrival followed neither of these models. Instead, traffic followed a more general model dubbed the “packet train,” which describes network traffic as a collection of packet streams traveling between pairs of nodes. A packet train consists of a number of packets travelling between a particular node pair.
This study examines the existence of packet trains on NSFNET, a high speed national backbone network. Train characteristics on NSFNET are not as striking as those found on the MIT local network; however, certain protocols exhibit quite strong train behavior given the great number of hosts communicating through the backbone.
Descriptions of the packet train model can be found in [3] and [5].
[1]
D C Feldmeier.
Empirical Analysis of a Token Ring Network
,
1984
.
[2]
Raj Jain,et al.
Packet Trains-Measurements and a New Model for Computer Network Traffic
,
1986,
IEEE J. Sel. Areas Commun..
[3]
Riccardo Gusella.
The Analysis of Diskless Workstation Traffic on an Ethernet
,
1987
.
[4]
Lawrence H. Landweber,et al.
Optimizing bulk data transfer performance: a packet train model
,
1988,
SIGCOMM 1988.
[5]
Raj Jain,et al.
Characteristics of Destination Address Locality in Computer Networks: A Comparison of Caching Schemes
,
1990,
Comput. Networks ISDN Syst..
[6]
Mahadev Satyanarayanan,et al.
IPwatch: a tool for monitoring network locality
,
1990,
OPSR.