Browse Prior Art Database

Transient Priority in a Control System with Distributed Control

IP.com Disclosure Number: IPCOM000123498D
Original Publication Date: 1998-Dec-01
Included in the Prior Art Database: 2005-Apr-04
Document File: 1 page(s) / 53K

Publishing Venue

IBM

Related People

Case, RB: AUTHOR [+2]

Abstract

In a control system with many users sharing a bottleneck resource, feedback on the state of the resource is supplied to all users. Better fairness and quicker convergence to fairness is achieved by applying a Transient Priority algorithm to the individual users' feedback control systems. The particular control system for which this was developed is a rate-based flow control enhancement to APPN/HPR. In HPR, the advantages of transient priority are: Transmission startup gets fast ramp up performance This makes more effective use of network resources for short connections. Short transmissions get high priority This improves observed performance for short-lived connections, and helps ensure fairness among sessions of different lengths.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 59% of the total text.

Transient Priority in a Control System with Distributed Control

   In a control system with many users sharing a bottleneck
resource, feedback on the state of the resource is supplied to all
users.  Better fairness and quicker convergence to fairness is
achieved by applying a Transient Priority algorithm to the
individual users' feedback control systems.  The particular control
system for which this was developed is a rate-based flow control
enhancement to APPN/HPR.  In HPR, the advantages of transient
priority are:
  Transmission startup gets fast ramp up performance
    This makes more effective use of network resources for
    short connections.
  Short transmissions get high priority
    This improves observed performance for short-lived
    connections, and helps ensure fairness among sessions
    of different lengths.

   To avoid creating large discontinuities in a system made
of many connections sharing a resource, new connections must start
out using a small fraction of the system's capacity.  Over time, a
connection receives feedback on the current state of the system and
adapts appropriately.

   Since new connections must "start slow", short connections
may never get to their fair share.  Longer connections naturally get
a higher share or "priority".  Also, when a new connection starts to
share a network containing a resource that is already fully used, the
bottleneck resource becomes over committed.  Feedback of the
overcommitted stat...