Browse Prior Art Database

Method of Making Collision Network Performance More Linear

IP.com Disclosure Number: IPCOM000107547D
Original Publication Date: 1992-Mar-01
Included in the Prior Art Database: 2005-Mar-21
Document File: 2 page(s) / 61K

Publishing Venue

IBM

Related People

Gittins, RS: AUTHOR

Abstract

This article proposes an alternative to the Ethernet collision handling. A collision is when an attempt to access the network is done at the same time as some other device attempts to access the network. The current method is to back off for a random interval in the event of a collision.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 58% of the total text.

Method of Making Collision Network Performance More Linear

       This article proposes an alternative to the Ethernet
collision handling.  A collision is when an attempt to access the
network is done at the same time as some other device attempts to
access the network. The current method is to back off for a random
interval in the event of a collision.

      The current method of dealing with collisions in an Ethernet
network has limitations.  First, it is not predictable, even at a
workstation.  This is probably of little importance.  Secondly, it
can result in network storms.  A network storm is when the collisions
are so frequent that the network becomes unusable, requiring all
devices on the network to back off and wait for a long period of
time, and then retry network access.  As the load increases, the
probability of a network storm increases more rapidly.

      Since the requests are still not satisfied, this is repeated
until requests are satisfied, or the requester gives up.  The side
effect is that the Ethernet usage will oscillate from idle to 100%
utilized.

      Needless to say, this condition is not desirable, since it
limits use of the Ethernet, by increasing access time, or limits the
number of users that can be connected to the network at any given
time.
Solution:

      A solution to this problem is to create a queueing protocol
that can be followed by all devices attached to the network.  This
allows access on a first-come...