Multiplex communications – Data flow congestion prevention or control – Control of data admission to the network
Reexamination Certificate
1998-12-01
2002-12-03
Yao, Kwang Bin (Department: 2664)
Multiplex communications
Data flow congestion prevention or control
Control of data admission to the network
C370S395200
Reexamination Certificate
active
06490249
ABSTRACT:
BACKGROUND OF THE INVENTION
Packet networks that offer their users certain level of performance usually require a set of traffic management capabilities. Among this set of capabilities, admission control plays a key role in limiting the volume of the traffic admitted to the network hence facilitate the support of QoS (Quality of Service) by the network.
The admission control function is often performed based on some expectation of the user's traffic profile. This profile is supplied to the network at the set up phase in the form of a limited set of traffic characteristics such as peak rate, sustained rate, and burst size. The values assigned to this set of traffic characteristic are often based on the user's best guess or some default values supplied by the manufacturers or the service providers or both. Consequently these values may bear little resemblance to the actual user activity and traffic pattern. Furthermore, the user is seldom active all the time, sending continuous traffic stream at the above rates, which causes a further diversion between the actual and the nominal traffic values.
An admission control procedure that is solely based on the supplied traffic parameters is usually referred to as mathematical CAC (connection admission control). A mathematical CAC usually leads to a network that is either over-utilized, hence is not able to guarantee the QoS parameters to its users, or a network that is under-utilized, hence not able to maximize its revenue. In most cases the latter case would prevail because of the tendency of the user to assign conservative values to their traffic parameters. Service providers usually compensate for the over-allocation of resources by introducing an overbooking factor that allows better utilization of their network resources.
It is therefore desirable to implement CAC procedures that take into account the current network status based on actual measurements. Since CAC procedures are different for the different services, it is also necessary to dynamically partition the network resources among the different services and virtual networks (customer networks running on a common service provider infrastructure) supported by the network.
The text of the ensuing description makes reference to the following documents by identifying the numeral associated with the corresponding reference as found to the left hand side below.
[1] R. Braden et. Al., “Resource Reservation Protocol (RSVP)—Version 1 Functional Specifications”, RFC 2205, September 1997.
[2] The ATM Forumn, User Network Interface (UNI) Signaling Specifications, February 1996.
[3] B. Jamoussi et. al., “System and Method for a Connection Admission Control Scheme in a Multi-Service ATM Networks”, Patent application, September 1997.
[4] B. Jamoussi, et. al., “Performance Evaluation of Connection Admission Control Techniques in ATM Networks”, IEEE Globecom Conference 1996.
[5] P. White and J. Crowcroft, “The Integrated Services in the Internet: State of the ART”, IEEE Proceedings, December 1997.
[6] ATM Forum, P-NNI Specification, 1995.
The CAC function plays an essential role for networks supporting some level of service guarantee for its customers. The first step of the CAC procedure is the reception of the connection establishment message containing the necessary information (traffic and QoS parameters) for the network to execute the mathematical CAC. The outcome of this step is the computation of the EBR (equivalent bit rate).
The EBR is combined in a certain fashion with the network measurements in order to formulate the admission criterion, taking into account the bandwidth assigned to the particular service class. The development of an EBR algorithm requires the mapping of the traffic parameters to an adequate statistical model and assumptions regarding the architecture of the transmission scheduler of the node and the size of the available buffer.
For the CBR (constant bit rate) and premium services where the traffic peak rate is of paramount importance, the traffic parameter of interest to the EBR algorithm is usually the traffic peak rate. For VBR (variable bit rate) services, the traffic parameters of interest are the peak rate, the average rate, and the burst size.
For CBR-like services, the easiest way to compute the EBR is to set its value equal to the connection's peak rate. This is particularly true if the node is equipped with sufficient buffer and the cell delay variation tolerance of the input scheme is small. More elaborate schemes are also possible [3].
For VBR-like service, one EBR algorithm that has seen wide use is the EGH (extended Gibbens-Hunt) algorithm [4]. EGH algorithm assumes a FIFO (first-in-first-out) scheduler for the node and an aggregate buffer of size B. The EGH algorithm maps the input traffic parameters to a two-state Markov-fluid model where a source alternates between active and idle states with exponentially distributed periods with parameters &agr; and &bgr; respectively. While being active, traffic is assumed to be generated with rate &lgr; units/s. With those assumptions, the EBR computed using the EGH algorithm for ATM-based networks is given by
EBR
=
-
[
α
+
β
-
γ
λ
]
+
[
α
+
β
-
γ
λ
]
2
+
4
γ
β
λ
2
γ
Where &ggr;_=_−log(CLR)/B*Pr[W>0], and CLR is the QoS parameter in terms of cell loss rate, B is the buffer size available at the node, and P[W>0] is the probability that the combined rate of the active sources exceeds the link capacity (probability of a non-empty queue).
Obviously the EGH algorithm is a loss-based algorithm in the sense that it is only applicable when the loss rate is the QoS parameter of interest. In many cases delay is the parameter of interest and the CAC procedure should be performed in a way to guarantee an upper bound on the delay experienced by a flow. An example of such service is the guaranteed service proposed for the Internet [5]. For the delay-dominated services, the EBR could be computed from [5]
Delay
=
(
b
-
M
)
×
(
p
-
EBR
)
EBR
(
p
-
r
)
+
M
+
C
tot
EBR
+
D
tot
where r and b are the rate and the depth of a token bucket characterizing the envelope of the flow, M is the maximum packet size, and C
tot
and D
tot
are error terms related to the scheduler and the finite packet sizes that are being dealt with.
An essential requirement of a signaling protocol used is its ability to provide, in addition to other information, the capability to carry the traffic and the QoS parameters requested by the user.
Instances of signaling protocols that satisfy those requirements are the RSVP protocol [1] for the Intemet (IP-based networks) and UNI (user-network interface) and P-NNI signaling [2] for ATM (asynchronous transfer mode) networks. In RSVP signaling, traffic parameters are carried in the T_spec field of the protocol and the results of the reservation procedures are carried in the R_spec field of the protocol. In UNI and P-NNI signaling, both the traffic and QoS parameters in terms of loss ratio and delay bounds are carried as IE (information elements) of the signaling protocol.
An example of routing protocols is the P-NNI (private network-to-network interface) of the ATM Forum [6] where the source node gathers the link state information advertised by the routing protocol to perform what is referred to as GCAC (generic CAC). The main function of the GCAC is to verify that an end-to-end path with sufficient resources will likely exist before performing the actual CAC on a link-by-link basis.
SUMMARY OF THE INVENTION
This invention describes a method and procedures of a hybrid CAC function that combines both the mathematical and the measurement aspects of the traffic. CAC functions based on some form of measurements is usually referred to as measur
Aboul-Magd Osama S.
Rabie Sameh A.
Gibbons Del Deo Dolan Griffinger & Vecchione
Harper Kevin C.
Nortel Networks Limited
Yao Kwang Bin
LandOfFree
Adaptive connection admission control scheme for packet... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Adaptive connection admission control scheme for packet..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Adaptive connection admission control scheme for packet... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2985326