Browse Prior Art Database

Adaptive Bi-Modal Filter for Estimating Link Load in Communication Networks

IP.com Disclosure Number: IPCOM000036523D
Original Publication Date: 1989-Oct-01
Included in the Prior Art Database: 2005-Jan-29
Document File: 3 page(s) / 34K

Publishing Venue

IBM

Related People

Ahmadi, H: AUTHOR [+2]

Abstract

In this article a novel adaptive bi-modal filtering technique to estimate very accurately certain important traffic parameters (e.g., utilization) which are needed for adaptive route calculation in packet- switched networks is proposed. Since the traffic parameters are generally time varying, the filtering mechanism must distinguish between a real change and a temporary variation in traffic and be accurate a large percentage of time. This filtering technique is very robust and responsive to real variations at one hand, and less sensitive to temporary variations in traffic conditions. This objective is also desirable in order to limit the number of unnecessary updates and reports for route calculations.

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 52% of the total text.

Page 1 of 3

Adaptive Bi-Modal Filter for Estimating Link Load in Communication Networks

In this article a novel adaptive bi-modal filtering technique to estimate very accurately certain important traffic parameters (e.g., utilization) which are needed for adaptive route calculation in packet- switched networks is proposed. Since the traffic parameters are generally time varying, the filtering mechanism must distinguish between a real change and a temporary variation in traffic and be accurate a large percentage of time. This filtering technique is very robust and responsive to real variations at one hand, and less sensitive to temporary variations in traffic conditions. This objective is also desirable in order to limit the number of unnecessary updates and reports for route calculations.

Obviously, the performance of the network can be improved if routing decisions take into account the current traffic conditions. This way the traffic can be distributed evenly among various routes that require the same class of service. There are already some existing networks which incorporate network traffic dynamics in their route computation algorithms. Some examples are the ARPANET, TYMNET and CODEX networks. In these networks some traffic related parameters are monitored and measured periodically and then, an average value over each period is reported at the end of each measurement interval. While the averaging of measurement data over each time interval would give a first order approximation of real traffic parameters, using the average measured data alone - without considering dependencies from one time interval to another - for routing decision could give unstable and undesirable results. The measured data is only one sample realization of the stochastic process. Hence, using measured data alone could lead into erroneous inference about the actual traffic parameters. The proposed filtering algorithm traces a real change in traffic conditions and filters out temporary variations. This feature of the filter is very desirable in order to limit the number of unnecessary updates and reports for route calculations. Estimation Technique

In order to estimate traffic parameters of each link of the network, we assume that traffic on each link is continuously monitored and measurements are taken. These measurements could be packet arrival counts, packet lengths, utilization, queueing delays or queue lengths. We consider time is divided into fixed intervals, each of duration T seconds. Let gt and mt denote the arrival and service rates of packets on a link, respectively. The time period T is assumed to be small enough that gt and mt are constant in the sense that within that period gt = g and mt = m. Therefore, the...