Learning-based admission control adjustment in data networks

Details Learning-based admission control adjustment in data networks Learning-based admission control adjustment in data networks

1999-09-28

2004-09-14

Chin, Wellington (Department: 2664)

Multiplex communications

Data flow congestion prevention or control

C370S395430

Reexamination Certificate

active

06791941

ABSTRACT:

FIELD OF THE INVENTION
The present invention relates to high-speed data networks, such as Asynchronous Transfer Mode (ATM) networks. More particularly, the present invention relates to Admission Control for bandwidth management and congestion control in such networks. Still more particularly, the present invention relates to the use of Connection Admission Control (CAC) adjustments in ATM networks using network measurement data to further control and tune an analytical CAC system.
BACKGROUND OF THE INVENTION
In broadband integrated services networks, e.g., those using asynchronous transfer mode (ATM) systems and techniques, information is packetized in fixed length “cells” for statistical multiplexing with other traffic for transmission over high-bit-rate channels. Such networks are connection oriented, so a connection must be established before transmission begins. Moreover, these connections are usually subject to contracts between a network operator and users of the network. To ensure quality of service (QoS) consistent with these contracts, connection admission control (CAC) techniques are typically employed in management of such networks. Generally, CAC algorithms determine whether a new virtual channel connection should be admitted to the network based on network status—such as available resources, cell loss performance—and contract parameters (e.g., mean traffic rate and peak traffic rate). See generally, Dziong, Z.,
ATM Network Resource Management
, McGraw-Hill, 1997.
Because of the complex variety of connection types and services, and consequent difficulty in ascertaining complete and current information regarding the actual state of ATM networks, and because of possible adverse consequences of failing to honor QoS guarantees in customer contracts, many network operators have chosen to use CAC algorithms that are quite conservative. Most CAC algorithms are designed for worst-case source behavior. Moreover, analytical models applied in these algorithms are also conservative—to account for the difficulty in achieving exact modeling of the connection aggregate process. Such conservative approaches in many cases tend to offset statistical multiplexing gains and other system efficiencies available in ATM networks.
Some have proposed using actual network measurements (such as traffic level and cell-loss characteristics in light of corresponding QoS constraints) to adjust CAC mechanisms in an attempt to more fully use network resources. See, for example, Bensaou, B.; Lam, S. T. C.; Chu, H. and Tsang, D. H. K., “Estimation of the Cell Loss Ratio in ATM Networks with a Fuzzy System and Application to Measurement-Based Call Admission Control,”
IEEE/ACM Transactions on Networking
, VOL. 5, NO. 4 (August 1997), pp. 572-584; Gibbens, R. J., Kelly, F. P., and Key, P. B., “A decision-theoretic approach to call admission control in ATM networks,”
IEEE Journal on Selected Areas in Communication,
13(6):1101-1114 (1995); and Saito, H. “Dynamic call admission control in ATM networks,
IEEE Journal on selected Areas in Communication,
9(7):982-989 (1991).
Thus far however, attempts to use network operating measurements have proven difficult in network administration, especially in respect of their incorporation in CAC processes. A particular difficulty arises in some prior art CAC processes in efficiently treating operations in networks exhibiting a wide variety of traffic types with a concomitant variety of QoS constraints. High bandwidth efficiencies through CAC tuning have not been readily available without high precision measurements.
SUMMARY OF THE INVENTION
The present invention overcomes limitations of prior art CAC algorithms and achieves a technical advance, as described in connection with illustrative embodiments presented below.
In accordance with one aspect of the present invention, the concept of aggregate effective bandwidth, AEBW, is used to provide a useful approximation to required bandwidth for given levels and classes of network traffic. AEBW is used in deriving an allowed level of overbooking—expressed in terms of overbooking gain, &agr;
t
.
In accordance with another aspect of the present invention, a learning-based system and method for CAC model tuning is employed, typically in the context of a network operations system, to provide information to switches to control over-booking gain. This over-booking gain is derived from longer-term network behavior and is updated from time to time based on measurement values sent from individual switches to an operations system.
In operation, overbooking gain at individual switches is incremented in small steps to achieve higher link utilization, until a threshold level is reached reflecting an excess for a measured operating parameter—such as cell loss. When the threshold is reached, overbooking gain is decremented by a larger amount relative to the increments; and the incrementing-decrementing process is repeated. The size of the steps and timing of updates is typically based on longer-term statistics of network operations, and can be modulated by operator control.


REFERENCES:
patent: 5315586 (1994-05-01), Charvillat
patent: 5357507 (1994-10-01), Hughes et al.
patent: 5463620 (1995-10-01), Sriram
patent: 5570360 (1996-10-01), Klausmeier et al.
patent: 5583857 (1996-12-01), Soumiya et al.
patent: 5862126 (1999-01-01), Shah et al.
patent: 5872771 (1999-02-01), Park et al.
patent: 5878029 (1999-03-01), Hasegawa et al.
patent: 5881049 (1999-03-01), Beshai et al.
patent: 5917804 (1999-06-01), Shah et al.
patent: 5953338 (1999-09-01), Ma et al.
patent: 6097722 (2000-08-01), Graham et al.
patent: 6324166 (2001-11-01), Yokoyama et al.
patent: 6459681 (2002-10-01), Oliva
patent: 6490249 (2002-12-01), Aboul-Magd et al.
patent: 6625155 (2003-09-01), Dziong

Affiliated with

Also associated with

Rating

Learning-based admission control adjustment in data networks does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Learning-based admission control adjustment in data networks, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Learning-based admission control adjustment in data networks will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-3264768