Browse Prior Art Database

Delay Trajectory Controlled Switching

IP.com Disclosure Number: IPCOM000081692D
Original Publication Date: 1974-Jul-01
Included in the Prior Art Database: 2005-Feb-28
Document File: 2 page(s) / 40K

Publishing Venue

IBM

Abstract

The resources of a store and forward communications network are dynamically allocated in such a way that the performance seen by the users is optimized in a particular sense. By "network" one or more nodes and interconnections are intended, and by "resources" line, processor and memory capacity are intended.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 60% of the total text.

Page 1 of 2

Delay Trajectory Controlled Switching

The resources of a store and forward communications network are dynamically allocated in such a way that the performance seen by the users is optimized in a particular sense. By "network" one or more nodes and interconnections are intended, and by "resources" line, processor and memory capacity are intended.

The basic idea is to assign priorities to messages to order the queues, such that delays through a node (Fig. 1) (or through the network in the more general case) become more nearly constant over a particular path.

When the delay through a particular node is short, the buffer or message store in that node has a great deal of free capacity. This storage capacity is then used to slightly delay those messages which would by chance pass through the node or network with abnormally short delay.

90% of all messages passing through a network lie within an area as shown in Fig. 2. The possible spread in delay times, of course, increases with the number of nodes through which a message passes.

The reallocation of delay is obtained by ranking or ordering of the messages in the queues within each node, based on a comparison of the actual delay a message has experienced with the target delay at the particular node.

One efficient means of achieving this is ranking according to a "GO-TIME" associated with each message. The following definition can be used: GO-TIME = (TIME OF BIRTH) + (TARGET DELAY) - (MESSAGE

LENGTH) (TRANSMISSION LINE...