Browse Prior Art Database

Adaptive Flow Balancing Congestion Control for Computer Communication Networks

IP.com Disclosure Number: IPCOM000049501D
Original Publication Date: 1982-Jun-01
Included in the Prior Art Database: 2005-Feb-09
Document File: 3 page(s) / 44K

Publishing Venue

IBM

Related People

Bharath-Kumar, K: AUTHOR [+2]

Abstract

A congestion control scheme for controlling buffer occupancy at the nodes of a communication subnetwork has been proposed. That mechanism is referred to as Flow Balancing Congestion Control (FBCC). The effectiveness of FBCC is documented in (2). The essential control parameter of FBCC is the threshold (THLD) which should be specified for every buffer pool associated with an out-going channel at every node.

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 54% of the total text.

Page 1 of 3

Adaptive Flow Balancing Congestion Control for Computer Communication Networks

A congestion control scheme for controlling buffer occupancy at the nodes of a communication subnetwork has been proposed. That mechanism is referred to as Flow Balancing Congestion Control (FBCC). The effectiveness of FBCC is documented in (2). The essential control parameter of FBCC is the threshold (THLD) which should be specified for every buffer pool associated with an out- going channel at every node.

An improvement of FBCC is described which enables the mechanism to find a proper value for THLD adaptively, according to the load of the system. The motivation for this improvement is as follows:

Basically, two parameters should be specified for FBCC: THLD and time-out (T.O.). Simulation studies indicate that performance of the mechanism is not sensitive to T.O. On the other hand, THLD is critical to performance (2), and its value should be chosen appropriately. Therefore. by determining THLD adaptively, FBCC becomes virtually parameterless.

The main performance measure considered is the histogram of buffer occupancy obtained by a long term observation of the buffer under control. Apart from the qualitative picture yielded by the histogram, several quantitative performance measures that are useful in determining the efficacy of the scheme can be identified. They are characterized below.
1. The probability of the buffer being zero is denoted as PO.

During congestion the introduction of control is intended

to keep the buffer occupancy small, without making a

large sacrifice in channel utilization (throughput).

This implies that PO should be kept small.
2. The moments of the occupancy histogram.

This provides a measure of the average delay and the variance

of the delay at the node. It is desirable that they be small

and controllable during periods of overload.

To consider the solution, see Fig. 1, which is a sketched histogram of the number of messages in the out-queue when FBCC is used and THLD is changed. Curve A in this figure is the result of THLD being too small. When THLD is too large, the histogram is shifted to the right (curve B). Curve C is obtained when a proper value for THLD is chosen.

The idea behind adaptive THLD is to shift the histogram far to the left. Since a change in THLD affects the probability of the queue being empty, which is another important measure, the decision to change.THLD...