Browse Prior Art Database

Real Time Bandwidth Adaptation Algorithm

IP.com Disclosure Number: IPCOM000103630D
Original Publication Date: 1993-Jan-01
Included in the Prior Art Database: 2005-Mar-18
Document File: 2 page(s) / 95K

Publishing Venue

IBM

Related People

Ahmadi, H: AUTHOR [+6]

Abstract

This article provides an algorithm for generating bandwidth updates in a bandwidth-managed network as the traffic characteristics of the source change in time.

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

Real Time Bandwidth Adaptation Algorithm

       This article provides an algorithm for generating
bandwidth updates in a bandwidth-managed network as the traffic
characteristics of the source change in time.

      High-speed networking protocols provide a quality of service,
and band-width guarantee to every transport connection established
across the network.  After a transport connection with specific
bandwidth requirements is set up with the necessary bandwidth
reservation along the path, the congestion control mechanism ensures
that the traffic going into the network stays within the allocated
bandwidth.

      The actual traffic characteristics of a connection can, in
general, vary during the life of a connection.  Here the arrival
process is represented by an on/off model.  When on, the source
generates traffic at rate R.  On and off times are independent from
each other and exponentially distributed.  b denotes the mean on
time, and p is the probability that the source is on.  The peak rate
R represents the characteristics of the port interface and is
constant.  However, p and b may change with time.  Therefore, these
two parameters are estimated and, based on the changes in these
parameters, bandwidth adjustment request messages are issued.  When
granted, the congestion control parameters are reset to the new
values based on the estimated user parameters.

      User traffic is continuously monitored and measurements are
taken on a fixed interval basis every T seconds.  The measurement
interval T should be at least the maximum of two time periods: (i) T
should be long enough to observe a large number of on/off periods and
(ii) T should be no shorter than the time it takes for an update
request to be processed/granted by the network.

      Let 0t be the vector of measured quantities at time t and let
          (pt,bt) = f(0s,0<s<t)
be the running estimates of p and b after the tth measurement
interval.  The function f denotes the estimator.  The measurement
vector 0 and the estimator f is beyond the scope of this disclosure.

      The alg...