Bandwidth allocation based on real time calculations using the convolution approach

Abstract

This paper focusses on one of the methods for bandwidth allocation in an ATM network: the convolution approach. The convolution approach permits an accurate study of the system load in statistical terms by accumulated calculations, since probabilistic results of the bandwidth allocation can be obtained. Nevertheless, the convolution approach has a high cost in terms of calculation and storage requirements. This aspect makes real-time calculations difficult, so many authors do not consider this approach. With the aim of reducing the cost we propose to use the multinomial distribution function: the Enhanced Convolution Approach (ECA). This permits direct computation of the associated probabilities of the instantaneous bandwidth requirements and makes a simple deconvolution process possible. The ECA is used in Connection Acceptance Control, and some results are presented.Our study focusses on the stationary model. In this case the effect of statistical multiplexing is the dominant factor, and it considers that all the excess cells are lost when the instantaneous rate is greater than the bandwidth provided by the link.Several limitations have been found in the previous studies: a) only sources of two-state, ON-OFF sources are presented; b) accurate evaluations have been simplified in order to reduce the complexity of calculations and the memory, and in [7] the convolution approach is applied only as binomial distribution over homogeneous sources; c) no distinctions have been made applying different QOS to the individual connections; and d) all models studied describe the behavior of the sources without considering the interactions in the network. The impact of the sources is studied in an individual manner (fluid-flow model), or a statistical evaluation is applied to a set of sources (stationary model).Our study has been carried out under the following premises:The basic objective of a bandwidth management and traffic control strategy in ATh4 network is to allow for high utilization of network resources, whle sustaining an acceptable QOS for all connections. In ATM, all the existing connections on a link are statistically multiplexed. Note that statistical multiplexing is efficient and allows more calls to enter the network than in the case of peak rate allocation, since it exploits the small probability that a larger number of calls will be active simultaneously [9] and [17].It is possible to assign an equivalent bandwidth (effective bandwidth for some authors) to each source which reflects its characteristics. Several network traffic control functions such as congestion control and routing depend on the characterization of the equivalent bandwidth of individual connections and the resulting load on networks links. A major challenge is to provide traffk control hctions in real-time. This normally involves a reduction in the complexity, and of course the accuracy, of the evaluation models.The analytical approaches to the blocked probabilities are due to [lo]. Other graphcal results for source load are presen...

Resource dimensioning aspect of heterogeneous traffic with different service requirements: integration versus segregation

“…In [6] we propose the New Convolution Approach to calculate the bandwidth required by dlfferent traf€ic types to achieve a certain QOS. When using this results, an iterative algorithm determines the sufficient and minimum capacity of a VP so that all the individual QOS are satisfied whenever different traflic classes are multiplexed together on the VP.…”

Study and Analysis of Cellular Wireless Networks with Multiclass Traffic

“…Now, it can be observed that N(t) = f (t, t , t3 t If n calls have arrived and each has the probability p of It can be noted that F(n+I) is the convolution [20] of F(n) and independently not completing by time t, then a sequence of F. Assuming f be the density function of F, it can be n Bernoulli [17] trials is obtained. Thus, …”