Dismiss
InnovationQ will be updated on Sunday, Oct. 22, from 10am ET - noon. You may experience brief service interruptions during that time.
Browse Prior Art Database

Adaptive Rate Control in a Start/Stop Flow System

IP.com Disclosure Number: IPCOM000111240D
Original Publication Date: 1994-Feb-01
Included in the Prior Art Database: 2005-Mar-26
Document File: 2 page(s) / 63K

Publishing Venue

IBM

Related People

Relyea, RJ: AUTHOR

Abstract

A method is disclosed for calculating rate based information from only start/stop signals. Given a rate based producer of data which has some mechanism of feedback for speeding up or slowing down, and a bursty consumer which consumes at some average rate, and only provides start and stop feedback based on high and low water marks in a queue. The producer would like to match the rate of the consumer to limit the number of full/empty messages getting sent. The producer would select a rate (R) which is somewhat greater than the consumers rate (Rc). The producer times how long it takes to reach the high water mark from the low water mark (tHWM - tLWM).

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 52% of the total text.

Adaptive Rate Control in a Start/Stop Flow System

      A method is disclosed for calculating rate based information
from only start/stop signals.  Given a rate based producer of data
which has some mechanism of feedback for speeding up or slowing down,
and a bursty consumer which consumes at some average rate, and only
provides start and stop feedback based on high and low water marks in
a queue.  The producer would like to match the rate of the consumer
to limit the number of full/empty messages getting sent.  The
producer would select a rate (R) which is somewhat greater than the
consumers rate (Rc).  The producer times how long it takes to reach
the high water mark from the low water mark (tHWM - tLWM).  At this
point, Rc can be calculated exactly as Rc = R - (SHWM - SLWM/tHWM -
tLWM) where SHWM is the size of the queue at the high water mark and
SLWM is the size of the queue at the low water mark.

      In a real system, there will be some error in tHWM, and tLWM,
and R and Rc may not be truly constant values.  These problems can be
solved by purposefully select R such that it is slightly faster than
Rc when the low water mark is reached, and then select R such that it
is slightly slower than Rc in the high water mark.  The difference
the producer selects between R and Rc, deltaR is based on the
variance of Rc and the uncertainty in tHWM-tLWM.  As R approaches Rc,
the difference between tHWM and tLWM increases.  Since the
uncertainty of these values is usually based on the res...