Diffusion Models for Computer-Communication Networks

Abstract

The diffusion theory is used to model a computer-communication network in which messages flow from one computer center to another. The idea is to approximate the various queueing processes that occur in the system (of discrete nature themselves) as continuous-state processes. The messages waiting at the queues to be transmitted are considered of small duration so that in the limit the flow of messages is continuous. With these ideas a general model for routing of messages in a computer network is established and expressions for the diffusion parameters (drift and covariance per unit time) are derived in terms of the network traffic. The mean length of the queues can thus be calculated and procedures to minimize the system overall queue size may be applied. Examples for simple networks are shown. One of them corresponds to a load-sharing computer system and it is indicated how the general diffusion methods derived earlier for message routing, can be used. Finally, a comparison is made between the expressions obtained by diffusion techniques and those corresponding to the classical exponential M/M/l queue.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
May 01, 1976
Accession Number
ADA031551

Entities

People

  • Jose H. Barbera

Organizations

  • Massachusetts Institute of Technology

Tags

DTIC Thesaurus Topics

  • Algorithms
  • Channel Capacity
  • Communication Networks
  • Computer Communications
  • Computer Networks
  • Computer Science
  • Covariance
  • Differential Equations
  • Diffusion
  • Diffusion Theory
  • Equations
  • Information Processing
  • Mathematical Analysis
  • Probability
  • Random Variables
  • Random Walk
  • Stochastic Processes

Fields of Study

  • Computer science
  • Mathematics

Readers

  • Computer Networking
  • Mathematical Modeling and Probability Theory.

Technology Areas

  • AI & ML
  • AI & ML - Bayesian Inference
  • AI & ML - Machine Learning Algorithms