Network Traffic Engineering (eBook)

Stochastic Models and Applications

Andrea Baiocchi (Autor)

eBook Download: EPUB

2020
John Wiley & Sons (Verlag)
978-1-119-63251-1 (ISBN)

Infos

A comprehensive guide to the concepts and applications of queuing theory and traffic theory 

Network Traffic Engineering: Models and Applications provides an advanced level queuing theory guide for students with a strong mathematical background who are interested in analytic modeling and performance assessment of communication networks. 

The text begins with the basics of queueing theory before moving on to more advanced levels. The topics covered in the book are derived from the most cutting-edge research, project development, teaching activity, and discussions on the subject. They include applications of queuing and traffic theory in: 

• LTE networks 
• Wi-Fi networks 
• Ad-hoc networks 
• Automated vehicles 
• Congestion control on the Internet 

The distinguished author seeks to show how insight into practical and real-world problems can be gained by means of quantitative modeling. Perfect for graduate students of computer engineering, computer science, telecommunication engineering, and electrical engineering, Network Traffic Engineering offers a supremely practical approach to a rapidly developing field of study and industry.  

ANDREA BAIOCCHI, PhD, is a Full Professor in the Department of Information Engineering, Electronics and Telecommunications of the University of Roma 'La Sapienza'. He has published over 160 papers on international journals and conference proceedings. He has participated to the Technical Program Committees of more than seventy international conferences. He served in the editorial board of the telecommunications technical journal published by Telecom Italia (currently TIM) for ten years.

Textauszug

1
Introduction

Quelli che s'innamoran di pratica sanza scienza son come il nocchiere, ch'entra in navilio sanza timone o bussola, che mai ha certezza dove si vada.1

Leonardo da Vinci

1.1 Network Traffic Engineering: What, Why, How

Engineering is the application of scientific principles and results to the design and optimization of machine, processes, systems. The typical approach of engineers consists of understanding objectives and requirements, abstracting a model of the system to be designed, defining a solution approach and testing for its suitability, i.e., checking if relevant performance metrics meet the prescribed requirements.

A key point is the ability of deriving a simplified model from a description of the system or function to be designed. The model should be simple enough to lend itself to analysis and provide understanding of performance trade‐offs, yet it should not miss any feature having significant impact on the relevant performance indicators.

Optimization of the model is a second key step. This can be often stated as a constrained optimization problem, where constraints come from performance requirements, costs, physical limits of the system.

The entire modeling and design process can be conceived as a double loop (see Figure 1.1). First, comparison with simulations or experimental measurements leads to the refinement of the model, so that it can reliably match the relevant dynamics of the system to be modeled. Once the model is assessed, it is used to refine the system design and to pursue its optimization, according to the results of the analysis, leading to new (hopefully better) performance results.

Figure 1.1 Scheme of system modeling and design process: from system observation and description, to model definition and refinement, based on comparison with simulations or experimental measurements (lower loop), then model usage for system dimensioning and optimization, according to an iterative refinement process based on performance results checking (upper loop).

The very concise sketch of the engineering approach to problem solving is a general one. Traffic engineering refers to the design and optimization of a class of systems and processes: networked service systems.

Let us examine the keywords one by one.

Service system is an abstraction of any physical or logical function under Quality of Service (QoS) constraints. This is where the essence of service is.

Networked refers to the fact that multiple interconnected systems carry out the assigned task(s). For that purpose, “traffic” moves from one service system to another one, according to the topology of the interconnection and subject to the capacity of the network. We use terms in an informal way in this introductory section. So, by network capacity we mean the capability of the network to transfer resources (e.g., information, goods, vehicles) depending on the kind of service, hence network, we are considering (e.g., communication network, logistic network, transportation network) to provide service to users' demand.

Traffic can be defined informally as the stochastic process describing the users' service demand, as regards both time of demand submission to the system (arrival) and duration of service. Users of the service system (e.g., applications, persons, machines) require the service system to carry out its tasks to meet their service demand. Times when service demand is submitted to the system as well as the amount of work required to meet the specific demand can be characterized as random variables. Hence, traffic engineering is intimately connected with probability and stochastic processes theory and its applications, a prominent position being reserved to queueing theory. That is the preferential “language” of traffic engineering, even if also other mathematical tools are often used (e.g., fluid approximation theory, optimization theory, game theory, to mention a few).

Performance evaluation is at the heart of traffic engineering. A service system encompasses three major aspects: (i) users' traffic demand; (ii) serving capability and resources provided by the system; and (iii) QoS constraints. The aim of traffic engineering is the design of the service system to meet the expected users' demand under the prescribed quality constraints. Minimization of cost, both capital expenditure to set up the system and operational costs, is of paramount importance to the system provider. This is usually in conflict with meeting an assigned level of QoS, which is instead of primary relevance to the system users. Trading off costs for QoS, given the users' demand, is the core “business' of traffic engineering. Conversely, estimating the admissible demand for the desired level of QoS, given the available resources and the way the system is designed, is another key task of traffic engineering, leading to the definition of algorithms and procedures to rule the access of users to the system resources and to manage those resources (priority, scheduling, flow control, congestion control, multiple access).

The reason why such a discipline has been developed and has grown as a recognized field is that no ‘free” resource is given in any service system. Hence, rational design of what resources to use, how much of them, and how to use them, still providing a “useful” service (i.e., meeting a specified QoS level) is key to making design of service systems viable from a technical‐economic point of view. This is why the design of service systems calls for suitable quantitative methods, able to provide predictions of key performance indicators.

Since traffic engineering is based on system modeling and abstractions, it has long been recognized that many different technical fields give rise to networked service systems that lend themselves to common models, independent of details of the specific technology or application field. Mathematical tools have been developed that can be used across many different application areas to a very large spectrum of systems. To mention some of them, communication networks, computing systems, transportation networks, logistic networks, power grid networks, production processes, all can be cast into the service system abstraction and therefore be designed by resorting to network traffic engineering tools. Each example application domain is itself a highly structured and complex system, encompassing a huge variety of physical resources and processing logic (we could refer to them as “hardware” and “software,” borrowing a classic terminology of information technologies).

Networked service systems can be modeled and analyzed by using different approaches. More in‐depth, analysis, dimensioning, and optimization of service systems can be faced along three main lines:

1. 1. Analytical models
2. 2. Simulations
3. 3. Experiments

Analytical models provide a mathematical description of the system that yields to tractable analysis (closed formulas) or, more often, to numerical investigation. This is the most powerful approach for a quick and nontrivial understanding of the performance trade‐offs, to gauge stability margins of the system, to assess the impact of key system parameters on performance, to provide a setting for stating optimization problems. While producing an effective analytical model requires hard study and a bit of talent to strike the best balance between simplified assumptions and a representative model, the time and computational effort required to use an analytical model make it the least costly among the three approaches listed above. The real difficulty of an analytical model is not really in solving the model once stated (books are there just to provide a guide for that purpose). It is rather the ability “to make things simpler, but not easier,” to say it with Albert Einstein's words. The art of modeling consists in making all sort of assumptions leading to the simplest model that still captures the aspects that are decisive to give a sensible answer to questions on the system. A fluid model that disregards completely the discrete nature of packet traffic in a communication network, such as the Internet, can be perfectly acceptable when we set out to study algorithms for congestion control, whereas it is definitely inadequate if we are interested in characterizing the delay jitter of a packet voice multiplexer or the collision probability of a random access protocol.

Among analytical models, a major role is played by stochastic process theory and queueing theory. The most useful class of stochastic process for service system analysis relates to Markov chains. The success of stochastic process theory and queueing theory as tools for network traffic engineering motivates the space devoted to them in this book.

Analytical models provide answers to basic questions in a quick and lightweight way. We can gain valuable insight on the system performance cheaply. When it comes to assessing second‐order effects or we need to relax assumptions on the system model in a way that does not yield to analytical tractability any more, computer simulation is often a valid approach.

Computer simulation for network traffic engineering purposes amounts to defining a detailed operational model of the system and reproducing all processes involved in this detailed model by means of a computer program. This is obviously not a real‐life system; it is,...

Produkt-Details

• Erscheint lt. Verlag: 24.7.2020
• Sprache: englisch
• Themenwelt: Mathematik / Informatik ► Informatik ► Netzwerke
• Themenwelt: Technik ► Elektrotechnik / Energietechnik
• Themenwelt: Technik ► Nachrichtentechnik
• Schlagworte: Client/Server-Technologie • Client/Server Technologies • Communication Technology - Networks • Computer Science • congestion control in the Internet • Electrical & Electronics Engineering • Elektrotechnik u. Elektronik • engineering applications of traffic and queuing theory • Informatik • Kommunikationsnetz • Kommunikationsnetze • LTE • multiple random access • Netzwerk • performance evaluation of networked service systems • performance modeling • Quality of Service • Systems Engineering & Management • Systemtechnik • Systemtechnik u. -management • Traffic theory • Wi-Fi
• ISBN-10: 1-119-63251-X / 111963251X
• ISBN-13: 978-1-119-63251-1 / 9781119632511