Connection admission control for connection orientated networks

Details Connection admission control for connection orientated networks Connection admission control for connection orientated networks

1998-12-16

2003-12-16

Nguyen, Chau (Department: 2663)

Multiplex communications

Data flow congestion prevention or control

Control of data admission to the network

C370S235000, C370S395200

Reexamination Certificate

active

06665264

ABSTRACT:

FIELD OF THE INVENTION
The present invention relates to the control of call acceptance in a communications network and in particular, an Asynchronous Transfer Mode (ATM) network.
BACKGROUND TO THE INVENTION
ATM is widely perceived as the next generation of high speed networking technology. ATM will integrate various electronic communications media that are used today into a multi-service network platform. ATM is a connection oriented network which implies that when requested, a connection is set up among a number of network termination points. At connection request time, a decision must be made whether the network can accept the new connection and satisfy its Quality of Service (QoS) parameters. The network also has to ensure that the existing connections in the network will not suffer degradation of QoS. Connection Admission Control (CAC) is the function in access ATM switches which performs this task. CAC is a component of every ATM switch that is capable of resource management. Normally it is positioned in the ATM Layer inside each and every ATM component that has authority over a physical link connected to it.
Previous research in the area of CAC for ATM networks has been concentrated in defining optimal statistical functions which make an efficient decision for one particular connection request. Typically, these solutions presume elaborate statistical traffic models characterizing a traffic source and are often computationally intensive. CAC is normally considered to be a control function inside an ATM switch with little other knowledge beyond the scope of the switch.
FIG. 1
shows how the connection control algorithm is normally presented. A source informs the CAC
2
of its need of a connection and the traffic characteristics it is going to use. These parameters are fed into a declarative part
3
which communicates with an estimation part
4
to get a clear picture of the strain that will be put on the network. The box
2
labelled “Connection Admission Control” shows the boundaries of the architectural specification of CAC. Many adaptive and intelligent algorithms have been proposed that need network measurements. However, since CAC is only active during the negotiation of the traffic contract no room has been left in the architectural specification for this behaviour.
A multiplex of voice, video and data connections appears to the network as a stream of cells sharing the same physical channel. Knowing the traffic descriptors and the QoS requirements, the CAC function must determine the amount of resources needed to achieve the traffic contract. CAC manages just one resource: network bandwidth. The simplest form of CAC is peak rate allocation. The algorithm for this ensures that the sum of the peak rates for each connection is less than the maximum utilization level—as a percentage—times the network capacity. A new connection request will be accepted or rejected solely on this criterion. The algorithm guarantees that no burst scale congestion occurs. However, cell scale congestion may still occur due to the discrete nature of the traffic at the cell level. This will lead to cell loss if no buffering is provided. Using appropriate queuing techniques the buffers can be dimensioned to guarantee less than a specified cell loss probability for a specified switch utilization.
Another form of CAC is an effective bandwidth scheme which preallocates an amount of bandwidth to each connection type (stored in a look-up table) that denotes the minimum amount of bandwidth needed for the connection to reach its QoS contract. The question of whether or not a new connection can be allowed on the network can then simply be answered by accumulating all effective bandwidths already in play on the link, adding the effective bandwidth of the new connection and comparing it to the link capacity. The problem with this approach is that the allocation of effective bandwidth tends to be somewhat conservative in order to ensure that the QoS contract is maintained at all times, irrespective of the real usage of bandwidth. This effectively restricts the number of connections that be accepted. Another, more computationally complex effective bandwidth scheme is one which calculates an effective bandwidth for each individual connection request on the basis of a number of traffic parameters specified by a source and by assuming certain mathematical traffic models. Statistical multiplexing techniques are used to determine the effective bandwidth, which will be a value between the peak and mean rates of the connection declared by the source. This technique is discussed in detail in the paper by R Guerin, H Ahmadi and M Haghsineh, “Equivalent capacity and its application to bandwidth allocation in high speed networks”, IEEE Journal Selected Areas on Communications, 9(7); 968-981, September 1991. The main problem with this effective bandwidth approach is that QoS can only be met if the real source traffic parameters conform to the source declarations or if the source declarations can be enforced by a policing mechanism.
Neural networks and fuzzy logic have also been proposed as CAC algorithms. These concepts attempt to predict the statistical behaviour of the multiplexed sources and use this to predict the cell loss rate. The decision of whether to reject or accept an incoming connection can be made by comparing this prediction of cell loss rate to the goal value. Neural networks and fuzzy logic can be implemented in hardware, are capable of learning (adaptive fuzzy logic) and can run with incomplete data. These attributes make them particularly well suited to connection admission control, although they are difficult to train because of the diverse ATM traffic characteristics and QoS requirements.
Without a CAC algorithm the network has no preventive method for traffic congestion. Yet the question of the ideal algorithm for CAC is a difficult one to answer. The algorithm needs to make educated decisions on traffic flows that may vary greatly and have poorly defined characteristics. Furthermore CAC has to find a balance between the needs of the network operator and the users; a CAC algorithm should allow the network to run efficiently but should also guarantee a quality of service agreed with the user in the traffic contract. To further complicate the issue some connections may need guarantees about cell inter-arrival times.
The conventional schemes discussed above tackle this complexity in two ways: by assuming that the CAC function should only have a microscopic view of the network and by making assumptions about traffic behaviour. However, ATM has been designed and branded from the start as being a flexible multi-service networking architecture and therefore these existing algorithms make inappropriate assumptions and ultimately cannot be guaranteed to make efficient use of bandwidth.
SUMMARY OF THE INVENTION
According to a first aspect of the present invention, a method of providing control of call acceptance in a node of a connection orientated communications network comprises the step of periodically modifying a connection admission control function of the node in dependence on a measurement of bandwidth utilization of the node over an interval encompassing a number of call connections, wherein the connection admission control function of the node implements an effective bandwidth scheme which allocates an amount of bandwidth to each of a number of connection types and one or more higher layer connection admission control functions are responsive to traffic fluctuations to modify effective bandwidths for connection types stored by the node in a predetermined manner.
The measurement of bandwidth utilization may be made by the node itself. However, it is preferred that the measurement is made at a network management level since this simplifies the design of the node and makes use of the inherent computational resources of the overlying management network.
Preferably, the measurement is also made over a second, longer, interval as part of a third layer CAC function. Additional higher level CAC layers m

Affiliated with

Also associated with

Rating

Connection admission control for connection orientated 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 Connection admission control for connection orientated networks, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Connection admission control for connection orientated networks will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-3169328