Today, the importance of providing quality of service (QoS) in the network for various types of real-time applications is widely accepted. Examples for such services are video communications and distributed co-working scenarios. Various mechanisms to provide or even to guarantee some minimum QoS values have been developed and others are current work in progress. Typically, the problem for an application and an end user is to estimate the currently available network quality. Even if QoS mechanisms are employed on all used parts of the network, it cannot be taken for granted that the prerequisites are met. Besides measurement mechanisms and analysis methods a metric is required allowing a numerical comparison of the available service quality and the demands of the particular application. This paper introduces a new calculation method which allows one to use measured QoS values and the application requirements to compute a numerical representation for the service quality. Additionally, the result of the computation enables a direct comparison of different communication paths for the same kind of service.
Currently, traffic engineering means circuit or virtual circuit switched technology even in packet switched networks. The need for a simpler and more suitable device for IP networks is getting stronger. In this paper, we present a new architecture for IP layer traffic engineering called Core State Limited Load Sharing (CSLLS). Our solution is highly tunable between low complexity—low granularity and high complexity-high granularity operation thus it can fit the requirements of a broad range of networks and services. CSLLS is a Differentiated Services conform framework, and was developed with an eye on efficiency. Hence, it realizes thrifty traffic engineering. We believe that, regarding the benefits of our proposal, CSLLS could become a serious rival of MPLS in IP networks.
Capacity planning is indispensable for future Internet providing QoS. Accurate dimensioning is especially important when no per-flow signaling or control exists.In this paper, we address the problem of link dimensioning and Lable Switching Path (LSP) optimization for MPLS networks supporting DiffServ EF and BE traffic classes. The problem is formulated as an optimization problem, where the goal is to minimize the nonlinear total link cost, subject to the performance constraints of both expedited forward (EF) and best effort (BE) traffic classes. The variables to be determined are the routing of LSPs carrying both EF and BE traffic, and the discrete capacities of the links.We show that Lagrangean Relaxation and subgradient optimization methods can be used to effectively solve this difficult problem. Computational results show that the solution quality is verifiably good, while the running time remains reasonable on practical-sized networks. This is the first work on capacity planning for MPLS networks supporting multiple Diffserv service classes.
Traditional approaches to network provisioning assume availability of the reliable estimates for the expected demands. This assumption, however, oversimplifies many practical situations when some incomplete information on the expected demands is available, and proper utilization of this information may improve the network performance. In a case of traffic engineering the uncertainty in the expected demands may be a result of unexpected and sudden changes in the demand pattern when difficulty of determining the varying demand pattern and possible undesirable transient effects make continuous adjustment of the routing algorithm to varying demands difficult. Assuming that some incomplete information on the future demand is available, this paper proposes a framework for network provisioning by balancing risks of under and over provisioning of the network.
Multi-Protocol Label Switching (MPLS) provides ways to control the Label Switched Paths (LSPs) followed by traffic trunks in a network and thereby to better traffic engineer it. In this context, we look at the problem of organizing the mapping of LSPs in an optimal way throughout the network on the basis of a given objective function. This problem is highly combinatorial and makes dynamic and real-time features a difficult issue for any LSP routing scheme. For this reason, we propose a computationally efficient, though approximate, on-line scheme adapted to an incremental optimization of the network state. It is then applied to a seldom mentioned traffic engineering problem: the compromise between load-balancing and traffic minimization. It is expected that clever routing strategies to balance the network load will sometimes favor longer paths in order to avoid congestion, leading to an increase of the overall network utilization. This reasoning is confirmed by our study, and we show that an improvement in network management can be made by appropriately tuning this compromise.
This paper represents a first step in exploring the formation of connected topologies in ad-hoc networks built on the Bluetooth technology. Connectivity is the most basic requirement for any system aimed at allowing devices to communicate with each other and in this paper we illustrate that this seemingly innocuous goal gives rise to many significant challenges in the context of the Bluetooth technology. We start with a brief overview of Bluetooth and its operation and then identify some of the major problems the technology faces when used to build ad-hoc networks. The paper's contributions are in introducing basic algorithmic problems associated with building connected Bluetooth networks and in developing several possible solutions capable of generating "good" connected topologies.
In this paper, we review several network admission control (NAC) methods. We explain how the NAC budgets and the required link capacities can be dimensioned based on a traffic matrix, a desired blocking probability, and the routing. The objective of this work is the inversion of that process. Based on a traffic matrix, the routing, and given link capacities, the budgets are to be assigned such that their blocking probabilities are as low as possible. We present an algorithm for fair resource assignment and illustrate its effect on a single link. We extend this mechanism to entire networks, such that it is adaptable to all budget-based NAC approaches. The evaluation of our concept shows that it is most effective in real networking scenarios where heterogeneous traffic patterns occur.
The Internet topology at the Autonomous Systems level (AS graph) has a power-law degree distribution and a tier structure. In this paper, we introduce the Interactive Growth (IG) model based on the joint growth of new nodes and new links. This simple and dynamic model compares favorable with other Internet power-law topology generators because it not only closely resembles the degree distribution of the AS graph, but also accurately matches the hierarchical structure, which is measured by the recently reported rich-club phenomenon.
All service control nodes, for example SCPs, HLRs, and MSCs, are constructed as a server system with one or more servers that process incoming requests. Since service control nodes are sensitive to overload, admission control mechanisms are often implemented. In this paper, we model and analyse admission control mechanisms for server systems with control theoretic methods. We develop a non-linear stochastic control theoretic model of a GI/G/1-system. Further, we show that this model can be used to design admission control mechanisms that behave well in the "real" system, that is the queuing system. We develop two types of controllers for the system, one PI-controller and one RST-controller, both commonly used in automatic control. Also, we discuss some of the limitations of previous work in this area, in which only deterministic and linear models have been used.
We focus on the optimal size a data burst (DB) should have in an Optical Burst Switching (OBS) network in the single wavelength link case. The optimality is understood in the sense of maximizing throughput under the assumption the traffic is TCP controlled. In such networks, packets are assembled into bursts containing several IP packets (burstification). A trade-off takes place between large and small DBs. The former leads to a degradation of throughput due to loss synchronization, a well known problem in TCP controlled traffic, whereas the latter leads to overhead due to the guard-band intervals. To address the optimal DBs size problem, we use an estimate of throughput obtained through individual TCP connections sharing a common router. The effect on synchronization is thus taken into account. A classical optimization technique is then applied to the resulting goodput formula to determine the optimal size. Besides, we also examine Fiber Delay Lines (FDL) modeling aspects.
We develop an optimization framework for the network carrier to manage profit in a two-tier market environment, a retail market where bandwidth is provisioned to serve uncertain demand and a wholesale market where bandwidth is bought and sold as a commodity. Our model is built upon mean-risk analysis. We discuss the selection of a risk index that is consistent with a stochastic efficiency criteria. We conduct numerical studies to investigate the influence of the carrier's risk averseness on bandwidth management, and profit implications of the bandwidth wholesale market.
As the Internet has become a big business, its performance has become an important question. Several companies have started monitoring Internet performance, many ISPs now conduct internal performance measurements, and there are also independent organizations such as RIPE that conduct regular performance monitoring of ISPs. The exact form of measurements may vary, but typically they are based on active probing. A common result presented is an overall statistic combining the measurements over a whole network, and this is frequently used to rate, or rank ISP performance, and is sometimes used as a metric for overall network health. Of course, a single number cannot hope to validly represent all of the information gathered in these types of measurements, but nevertheless, this type of measure is used, and we should ensure that the methodology for compiling the measurements is robust, and appropriate. Most methods that have been used are based on the mean, but it is surprising how diverse the "mean" can be. There are several alternatives (e.g. arithmetic mean versus geometric mean), and each company applying these methods seems to use a different approach. This paper considers five approaches in detail and, explains which of these is the best, and why. In particular, we show that the geometric, and harmonic means, while appealing because of their robustness to outliers, are actually very poor statistics to use when combining Internet measurements, and can cause changes in apparent performance without any real change in the network performance.
After discussion of SLA issues the mathematical methodology - how to design the global SLA index on the basis of Linear Discriminant Functions - is developed. The experience of Verizon New York company is used as a best practice. Service quality monitoring is a useful base for the teletraffic community to join around for future packet switching network studies.
This paper studies the maximum stable throughput of FS-ALOHA, a random access algorithm for dynamic bandwidth reservation in access networks, subject to delay constraints. Requests that are not transmitted successfully before the maximum delay t(max) expires are dropped. We state that the algorithm is stable for a certain input rate A, if the dropping probability is below a predefined tolerance epsilon, e.g., epsilon = 10(-9). The matrix analytic method (MAM) is used to determine the maximum input rate for which the system is stable. Numerical examples for different parameter setting provide a useful insight on how to optimize the maximum stable throughput as a function Of t(max) the maximum delay, and epsilon, the drop tolerance.
We present an integrated packet/flow level model for WLAN performance analysis. It captures the statistical characteristics of the transmission of individual packets at the MAC layer, and includes the system dynamics due to the initiation and completion of data flow transfers. The processor sharing-based model is analytically tractable and yields a simple approximation for the expected flow transfer time. Extensive simulations show that the approximation is very accurate for a wide range of parameter settings.
We give an analysis on the throughput of WLANs, with a particular focus on the IEEE802.11 a standard. Our model is a very generic one, which consists of a single access point and a number of terminals under heterogeneous radio conditions, and takes into account both the effect of collisions at the MAC layer and that of transmission errors under imperfect radio links. It is shown that both the system and individual throughputs depend very much on the packet size, the number of transmitters, the transport protocol (TCP/UDP), in addition to the radio environment and the selected transmission mode. Our results help one gain a good insight into the system performance of IEEE802.11 a WLANs under various system conditions.
Multifractal behavior was recently observed in several traces of IP WAN traffic. This paper proposes a novel multifractal traffic model, which characterizes the joint process of packet arrivals and packet sizes. The construction of the traffic process is based on stochastic L-Systems, which were introduced by biologist A. Lindenmayer as a method to model plant growth. We work with a single L-System alphabet and production rule, where the alphabet is a set of pairs, and each pair element represents a packet arrival rate and a packet mean size. In this way, the traffic model is able to capture correlations between arrivals and sizes, leading to an accurate prediction of the queuing behavior. We provide a detailed comparison with a related multifractal model based on conservative cascades. Our results, that include applying the fitting procedure to real observed data with multifractal scaling behavior on both the packet arrival and packet size processes, show that our L-System based model can achieve excellent fitting performance in terms of first and second order statistics and queuing behavior.
In this paper we compare the performance of two low latency handoff protocols for MIPv4, Pre- and Post-Registration Handoff. These mechanisms proposed by the IETF aim at improving the performance of Hierarchical Mobile IP with respect to handoff latency and packet loss. We propose an analytical model to study the influence of various system parameters on the performance of the two protocols, followed by a comparison of the two schemes. We describe several handoff implementations over a wireless access based on the IEEE 802.11 standard and analyze them by means of an ns simulation.
The fast computation of blocking probabilities and the resulting capacity is one of the crucial tasks in the planning process of UNITS networks. The admission control in WCDMA networks is based on the momentary interference which includes both own-cell and other-cell interference. Since both interference terms are stochastic values we speak of soft blocking. The number of users in the system is not sufficient for deciding whether to accept a new call or not. Instead, it is blocked with a certain probability depending on the number of users in the system, the activity of the users, and the other-cell interference. In this paper we present a time-efficient algorithm to compute blocking probabilities in a WCDMA network operating with several services. Assuming Bernoulli activity and modelling the other-cell interference as a lognormal random variable the blocking probabilities are computed using an approximation based on the Kaufman-Roberts recursion.
Ill this paper, we develop three explicit analytical models for evaluating the request blocking probability of movie files in VoD systems under three server selection schemes. The first is exact and the other two are based on fixed point approximations. We show that the choice of server selection schemes call significantly affect tire blocking probability performance of the systems. We demonstrate by simulation that the models accurately predict the overall file request blocking probability in the systems and the blocking probability of the requests for single-copy files. Comparisons among the three server selection schemes are performed. Observations are presented and intuitively explained.
