Catálogo de publicaciones - libros

In the Internet, a variety of entities competes with each other. For example, coexisting overlay networks compete for physical network resources and they often disrupt each other. Service providers offer different services and compete for network resources and customers. If competitors could establish cooperative relationships, the collective performance can be improved and they can coexist peacefully and comfortably in a shared environment. In this paper, to understand the way that symbiosis emerges from direct and/or indirect interactions among coexisting entities in a shared environment, we adopt a mathematical model of symbiotic strains. Through thorough evaluations, we show that networks of different service rate and resource consumption can coexist by mutual interactions.

This article proposes a hierarchical architecture for an infra-structure-based overlay network delivering multicast services. Such an overlay network is an alternative to IP multicast when the latter cannot be timely deployed for technical or business reasons. To make the overlay network scalable in terms of coverage and traffic volume, we propose a hierarchical architecture in which edge overlay routers are responsible for handing traffic to/from the end users, whereas core overlay routers handle transit traffic and perform the bulk of packet replication. We compare the hierarchical architecture to a flat architecture in which (edge) overlay routers are connected in a full mesh. We develop an asymptotic analysis to quantify the cost (in terms of additional required switching) and benefits of the hierarchical architecture over the flat architecture.

We study the capacity allocation problem in service overlay networks (SON)s with state-dependent connection routing based on revenue maximization. We formulate the dimensioning problem as one in profit maximization and propose a novel model with several new features. In particular the proposed methodology employs an efficient approximation for state dependent routing that reduces the cardinality of the problem. Moreover, the new formulation also takes into account the concept of network shadow prices in the capacity allocation process to improve the efficacy of the solution scheme.

Supporting multicast at WDM layer is an important requirement for high bandwidth multicast applications emerging in IP over WDM optical networks. Light-splitting technique has been proposed to provide light-trees, i.e., multicast trees in an optical layer. Many recent studies have been focused to efficiently build and configure light-trees without existing light-paths for unicast traffic into consideration. In this paper we identify and explore the optimal design problem of multicast configuration for realistic and constrained WDM networks. In such a network, both unicast and multicast are supported, and WDM switches have limited light splitting capability. Using wavelength sharing among traffic demands of unicast and multicast, we build a hybrid virtual topology which exploits both light-trees and light-paths. By optimizing WDM resources in addition to resource sharing with unicast, we truly maximize the WDM layer capability and efficiently support multicast traffic demands. We validate the efficiency of our approach with extensive simulations.

In this paper we investigate the characteristics of network traffic via the cluster point process framework. It is found that the exact distributional properties of the arrival process within a flow is not very relevant at large time scales or low frequencies. We also show that heavy-tailed flow duration does not automatically imply long-range dependence at the IP layer. Rather, the number of packets per flow has to be heavy-tailed with infinite variance to give rise to long-range dependent IP traffic. Even then, long-range dependence is not guaranteed if the interarrival times within a flow are much smaller than the interarrival times of flows. In this scenario, the resulting traffic behaves like a short-range dependent heavy-tailed process. We also found that long-range dependent interflow times do not contribute to the spectrum of IP traffic at low frequencies.

This paper develops a new state-space model for long-range dependent (LRD) teletraffic. A key advantage of the state-space approach is that forecasts can be performed on-line via the Kalman predictor. The new model is a finite-dimensional (i. e., truncated) state-space representation of the FARIMA (fractional autoregressive integrated moving average) process. Furthermore, we investigate, via simulations, the multistep ahead forecasts obtained from the new model and compare them with those achieved by fitting high-order autoregressive (AR) models.

Long-range dependence (LRD) or second-order self-similarity has been found to be an ubiquitous feature of internet traffic. In addition, several traffic data sets have been shown to possess multifractal behavior. In this paper, we present an algorithm to generate traffic traces that match the LRD and multifractal properties of the parent trace. Our algorithm is based on the decorrelating properties of the discrete wavelet transform (DWT) and the power of stationary bootstrap algorithm.

To evaluate our algorithm we use multiple synthetic and real data sets and demonstrate its accuracy in providing a close match to the LRD, multifractal properties and queueing behavior of the parent data set.We compare our algorithm with the traditional fractional gaussian noise (FGN) model and the more recent multifractal wavelet model (MWM) and establish that it outperforms both these models in matching real data.

With the increasing popularity of multimedia applications, video data represents a large portion of the traffic in modern networks. Consequently, adequate models of video traffic, characterized by a high burstiness and a strong positive correlation, are very important for the performance evaluation of network architectures and protocols. This paper presents new models for traffic with persistent correlations based on the M/G/∞ process. We derive two new discrete distributions for the service time of the M/G/∞ queueing system, flexible enough to give rise to processes whose correlation structure is able to exhibit both Short-Range Dependence (SRD) and Long-Range Dependence (LRD). The corresponding simulation models are easy to initialize in steady state. Moreover, as the two distributions have subexponential decay, we can apply a highly efficient and flexible generator of synthetic traces of the resulting M/G/∞ processes.

Several simulation studies have shown that the performance of IEEE 802.11 DCF in an ad hoc scenario strongly depends on the coverage and interference radii. We state and solve an analytical model for an 802.11 DCF ad hoc network, with an interference radius larger than the coverage radius. The model is developed for the study of a split channel solution, where RTS/CTS signaling is conveyed via a separate, orthogonal channel with respect to data and ACK frames. By exploiting the model we can optimize the bandwidth split of the control and data channels. Further, we compare single channel, split channel and multi-channel solutions, thus highlighting that the simple split channel achieves most of the performance advantage potentially offered by a multi-channel 802.11 DCF.

As the use of WLANs based on IEEE 802.11 increases, the need for QoS becomes more obvious. The new IEEE 802.11e standard aims at providing QoS, but its contention-based medium access mechanism, EDCA, provides only service differentiation, i.e. soft QoS. In order to provide hard QoS, earlier we have proposed an extension called (EDCA/RR), which enhances EDCA by offering also hard QoS through resource reservation. In this paper, we extend EDCA/RR to cope with the hidden terminal problem, outline a solution for multi-hop scenarios, and compare the proposed scheme with EDCA.