Browse Prior Art Database

Distributed Label Assignment in Gbit/s LANs

IP.com Disclosure Number: IPCOM000108481D
Original Publication Date: 1992-Jun-01
Included in the Prior Art Database: 2005-Mar-22
Document File: 3 page(s) / 157K

Publishing Venue

IBM

Related People

Le Boudec, JY: AUTHOR [+3]

Abstract

Disclosed is a distributed method for assigning a label (i.e., a short, 1 to 4 bytes, and locally administrated address) to a node each time it inserts into a Gbit/s LAN employing ring or dual-bus topologies, such as BCMA or CRMA-II. The method is fully distributed, that means it does not require any centralized function. It is also safe against multiple label assignment. Such labels have been proposed (1,2,3) to permit rapid data removal from the shared medium by the destination node, thus allowing the reuse of the transmission capacity.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 41% of the total text.

Distributed Label Assignment in Gbit/s LANs

       Disclosed is a distributed method for assigning a label
(i.e., a short, 1 to 4 bytes, and locally administrated address) to a
node each time it inserts into a Gbit/s LAN employing ring or
dual-bus topologies, such as BCMA or CRMA-II.  The method is fully
distributed, that means it does not require any centralized function.
It is also safe against multiple label assignment.  Such labels have
been proposed (1,2,3) to permit rapid data removal from the shared
medium by the destination node, thus allowing the reuse of the
transmission capacity.

      The process of removing data is simple, in that the capacity is
somehow marked as free.  On slotted LANs this involves only the
changing of a single bit.  However, the decision to remove must be
made extremely fast so that this marking can take place as the data
passes by the node at the transmission speed of the LAN.  Since this
decision is based on a node's recognizing its address, the address
must be of simple form and known to the receiving hardware.  Such an
address is called a label.  This article concerns the assignment of
labels to nodes.
LANs with a Ring Topology

      As mentioned before, labels are introduced in order to allow a
node a rapid removal of data destined to it.  For this aim, the
inclusion of two label fields in the data format is proposed (2,3),
the removal and the destination labels.  The destination label tells
a node to receive the data when its label matches the destination
label; the removal label indicates which node is to remove the data
by marking the capacity as free or idle.  For data addressed to a
single node, the destination label and the removal label would be the
same.  For data addressed to a group of nodes, the destination label
would be a group label and the removal label would be the label of
the source node.

      In principle the label assignment method presented in the
following is a combination of the above described label addressing
scheme with the name management method of NETBIOS (Network Basic
Input/Output System) (4).

      A newly inserted node should create its own label.  The method
of its choice, whether random or otherwise, is unimportant.  However,
instead of broadcasting the label to all attached nodes, it will send
a notification message on the LAN with the created label included
into both the destination and removal labels of the message and start
a timer T1.  This addresses the notification message only to any node
on the LAN that is using that label and does not interrupt all nodes.
The notification message contains further a unique identifier to
distinguish it from other nodes that might start the procedure at the
same time and create the same label (though the probability for that
event is very small).  The unique identifier could be a permanent and
globally administered address or a pseudo-random number.

      If another node already uses t...