Distributed admission control

#The present invention relates generally to networks and, more particularly, to admission control in networks.

#Data experiences latency and jitter (variances in latency) as it travels through a network from a source to a destination. Latency may be caused by several characteristics of the network, such as the communications medium the data is transported over, processing of the data as it travels through the network, congestion in the network, and other factors. Some causes of latency are unavoidable, such as physical limits of transport media. An example of a physical limit is the speed of light in an optical transport medium. Depending on the application, some latency and jitter may be acceptable. For example, for some applications it may be acceptable to have a certain average latency, latency variance, or some combination of these. At some point, however, long latency and high jitter become unacceptable for most applications.

#FIG. 1 illustrates the relationship between offered traffic load and latency in a network. As the offered load of traffic to the network increases, latency also increases. This general relationship between latency and offered load is a fundamental property of networks. At some point the amount of latency becomes unacceptable, which is generally defined as the point of congestion. Congestion is denoted in FIG. 1 by the horizontal line labeled “congestion.” Below the congestion line congestion does not occur, which is characterized by traffic that has acceptable latency and/or jitter characteristics. Above the line, congestion occurs. Congestion is characterized by traffic having unacceptably high latency, jitter, or both. When there is almost no load, latency drops to the lowest delay the network is capable of.

#The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate several embodiments of the invention and together with the description, serve to explain the principles of the invention.

#FIG. 1 is a diagram that illustrates the relationship between load and latency in a network;

#FIG. 2 illustrates a network using distributed admission control consistent with the principles of the invention;

#Reference will now be made in detail to embodiments consistent with the principles of the invention, examples of which are illustrated in the accompanying drawings. The same reference numbers may be used in different drawings to refer to the same or like parts.

#Some types of network applications are more latency- and jitter-sensitive than others. For latency- and jitter-sensitive applications, it is common to have a service-level agreement (SLA) between the customer and the service provider that requires the service provider to deliver data, such as latency-sensitive data, within specific parameters. For example, the SLA may include maximum latency and jitter requirements, as well as other service level requirements.

#In accordance with the principles of the invention, latency- and jitter-insensitive data is allowed in the network without controls or checks. For latency- and jitter-sensitive data, however, distributed admission control is carried out. According to embodiments consistent with the principles of the invention, distributed admission control may be carried out by sending one or more reflector packets to a destination and analyzing acknowledgments received in response to the one or more packets. More particularly, in order for a first network client to set up a data transfer with a second network client, the first network client notifies an admission control facility (ACF) that it would like to initiate the transfer. Before initiating the transfer, the ACF performs admission control to determine the state of network resources between the first client and second client. For example, the ACF may determine whether network resources between the source and destination have low enough latency and jitter levels to meet the requirements of an SLA for the data to be transferred between the first and second network clients.