Architecture of Network Systems

Dimitrios Serpanos, Tilman Wolf

• Publisher: Morgan Kaufmann
• Year: 2011
• ISBN:
  9780080922829
• Topic: Ciencia de la computación
• Subtopic: Redes de computadoras

Architecture of Network Systems explains the practice and methodologies that will allow you to solve a broad range of problems in system design, including problems related to security, quality of service, performance, manageability, and more. Leading researchers Dimitrios Serpanos and Tilman Wolf develop architectures for all network sub-systems, bridging the gap between operation and VLSI.This book provides comprehensive coverage of the technical aspects of network systems, including system-on-chip technologies, embedded protocol processing and high-performance, and low-power design. It develops a functional approach to network system architecture based on the OSI reference model, which is useful for practitioners at every level. It also covers both fundamentals and the latest developments in network systems architecture, including network-on-chip, network processors, algorithms for lookup and classification, and network systems for the next-generation Internet.The book is recommended for practicing engineers designing the architecture of network systems and graduate students in computer engineering and computer science studying network system design.

• This is the first book to provide comprehensive coverage of the technical aspects of network systems, including processing systems, hardware technologies, memory managers, software routers, and more
• Develops a systematic approach to network architectures, based on the OSI reference model, that is useful for practitioners at every level
• Covers both the important basics and cutting-edge topics in network systems architecture, including Quality of Service and Security for mobile, real-time P2P services, Low-Power Requirements for Mobile Systems, and next generation Internet systems

Chapter 1. Architecture of network systems overview

Abstract:

Computer networks have become critical infrastructure on which we rely for personal, business, and government use. Network systems are the hardware and software components from which these networks are built; they determine what functionality a computer network can provide and what performance it can achieve. Network systems are present in the infrastructure of networks, such as switches, bridges, routers, and modems. Network systems represent a distinct area of embedded systems architecture. Network systems draw from concepts and technologies in computer networks, computer organization, and distributed computing. The diversity of these technical areas requires a thorough understanding of the relationship between these areas and how they influence network system design. The functionality of network systems is determined by network protocol specifications. A systematic approach to network systems architecture is followed, where designs need to meet protocol requirements based on the Open Systems Interconnection reference model, performance requirements for high-speed networking, and technological and implementation constraints.

Keywords: network systems, architecture, computer networks, Internet, network infrastructure, protocols, embedded systems, Open Systems Interconnection reference model, layered protocol stack, performance requirements

Computer networks have become critical infrastructure on which we rely for personal, business, and government use. Network systems are the hardware and software components from which these networks are built. Network systems determine what functionality a computer network can provide and what performance it can achieve. Due to this critical role, we believe it is important to study the architecture and operation of these network systems.

Network systems draw from concepts and technologies in computer networks, embedded systems, computer organization, and distributed computing. The convergence of these very diverse technical areas makes the study of network systems particularly exciting. This diversity also requires a thorough understanding of the relationship between these areas and how they influence network system design. We hope to provide these insights in this book.

Computer networks

The advances of transmission technology for more than two decades have brought significant changes in networking as well as computing. In the 1970s and 1980s, standard networks provided limited connectivity, achieving bandwidth in the order of kilobits per second (kbps) up to a few megabits per second (Mbps) for local area networks, where the maximum speed reached 10 to 16 Mbps. From the middle of the 1980s, the development and commercialization of high-speed links that provided bandwidth of several Mbps for point-to-point connectivity enabled development of a new generation of networks and protocols that enable communication at very high speeds, reaching today hundreds of gigabits per second (Gbps).

In parallel with the dramatic progress in transmission technology, in the last decade of the 20th century the Internet was commercialized, moving it from research use to commercial use. The need to provide Internet connectivity to end users at home and at work not only exploited the high-speed transmission technology that had been developed, but also led to significant progress in access technologies. This trend led to development of a wide range of access protocols to connect end users to the Internet through telephone lines, cable TV infrastructure, satellites, and so forth.

The deployment of high-speed links and networks, as well as the Internet, provided the infrastructure for the development of new computing paradigms, mainly network-centric computing. In this paradigm, newly developed system infrastructures are used to support computing and storage-intensive applications and services. An early characteristic example is the development of networks of workstations, a multiprocessor architecture that relies on high-speed connectivity among workstations. This multiprocessor model is a natural advance of traditional distributed systems, which connected autonomous computing systems; the single view of the network of workstations as one system, necessary for a multiprocessor, was enabled by the high-speed networks that had become available. This abstraction enabled the efficient management of distributed resources through appropriate computing models and enabled a unified view of the networked workstations to the users. In a different direction, the ability to provide access to data and computational resources over the Internet enabled a vast number of new services for users and customers of commercial enterprises. These services are based on the well-known client/server distributed computing model and include examples ranging from banking to news feeds and from video conferencing to digital libraries.

The provision of all these services and applications over networks, including the Internet, requires technological advances at two fronts: protocols and network systems. Network protocols define the methods and mechanisms necessary to achieve reliable communication between two parties (or more than two in the case of multicasting or broadcasting). For example, network protocols define methods with which data units are encoded for transmission, mechanisms to detect transmission errors, methods for retransmission of data in case they are lost or transmitted with errors, and methods for regulating the flow of information between communicating peers to ensure that the receiver is not flooded with incoming data. Importantly, network protocols do not define any aspect of the systems that execute these protocols in order to implement data communication. For example, protocols do not define the type of processors, their speed, the size of memory, or any other systemic characteristic of the devices that implement these protocols.

Network systems are the systems and subsystems that realize the implementation of network protocols. Network systems need to be designed to meet the functional requirements specified by protocols. They also need to meet the performance requirements determined by the ever-increasing speed of transmission links. This relationship between network systems and related areas is illustrated in Figure 1-1. The demands for executing protocols at high speed led to the need for advanced, sophisticated system architectures, component designs, and implementations. These network systems constitute the focus of this book.

Figure 1-1 Requirements and constraints for network system design.

Network systems represent a distinct area of embedded systems architecture. Network systems are embedded systems because they are embedded in autonomous systems and devices that have specific purposes. For example, network systems are present in the infrastructure of networks, such as in switches, bridges, routers, and modems. Importantly, network systems also include network adapters, which are used in general-purpose computing systems, as well as in special-purpose systems, such as mobile phones.

The importance of network systems is increasing continually, driven by the dramatic growth in network-centric services developed and deployed. The expansion of broadband services has led to the continuing exponential growth of Internet users and the increasing adoption of services on the go (e.g., mobile banking, mobile TV), as well as the increasing deployment of large, networked sensing and monitoring systems (e.g., transported goods containers, environment monitoring systems). Importantly, these services are differentiated from traditional data connectivity services because they also provide real-time communication for voice and video services. Therefore, network systems have become critical components of the overall network infrastructure.

The design of network systems is challenging not only due to the increasing requirements to execute several complex protocols, but also because of the need to achieve efficient protocol execution within the resource limitations of embedded systems (i.e., size, power). Thus, the architecture of network systems constitutes a significant area of embedded systems architecture.

Embedded systems

Embedded systems are special-purpose computing systems embedded in application environments or in other computing systems and provide specialized support. The decreasing cost of processing power, combined with the decreasing cost of memory and the ability to design low-cost systems on chip, has led to the development and deployment of embedded computing systems in a wide range of application environments. Examples include network adapters for computing systems and mobile phones, control systems for air conditioning, industrial systems, and cars, and surveillance systems. Embedded systems for networking include two types of systems required for end-to-end service provision: infrastructure (core network) systems and end systems. The first category includes all systems required for the core network to operate, such as switches, bridges, and routers, while the second category includes systems visible to the end users, such as mobile phones and modems.

The importance of embedded systems is continuously increasing considering the breadth of application fields where they are used. For a long time, embedded systems have been used in many critical application domains, such as avionics and traffic management systems. Their broad use illustrates the importance of embedded systems, especially when considering the potential effects of their failure. For example, a failure of an automatic pilot system or a failure of a car braking system can lead to significant loss of life; failure of an electric power system may lead to loss of life or, if not to that, to loss of quality of life; and failure of a production control system in a factory may lead to a significant loss of revenue. Our dependence on embedded systems requires development and adoption of new architectural and design techniques in order to meet the necessary performance requirements and to achieve the required dependability using their limited resources in terms of processing, memory, and power.

The importance of embedded systems has led to the emergence of a strong industry that develops and uses them. Their criticality for services on all fronts and for technological and thus economic growth has led to significant efforts to address the challenges placed by embedded systems development and deployment. One important effort is the ARTEMIS initiative of the European Commission [1]. This program started with a Strategic Research Agenda (SRA) [8] and has grown to a significant activity, including a strong industrial association, named ARTEMISIA, which conducts research and development in the area of embedded systems. Figure 1-2, a figure from the ARTEMIS SRA [8], shows one view of the embedded systems area organized by research domains and application contexts. In Figure 1-2, horizontal bars constitute technological areas involved in embedded systems development and vertical bars indicate application contexts where embedded systems are used and are expected to penetrate applications in the future. Considering the differentiated requirements of embedded systems adoption in different application areas, Figure 1-2 groups in application contexts the services and applications that have common characteristics; different application contexts have significant differences among them. For example, the application context of private spaces includes systems and services for the home environment, the car, and private environments in general, where comfort and safety are the highest priority, while the context of industrial systems focuses on safety-critical systems for industry, avionics, and others. Clearly, the organization and semantics of application contexts change as time progresses and new applications and services are developed. One can organize the vertical bars with different criteria, such as, for example, the industrial sectors involved in the development of embedded systems.

Figure 1-2 ARTEMIS research domains and application contexts.

Protocols and network systems

As noted earlier, the networking field has focused mostly on the development of protocols for communication among network nodes. Considering the high bit error rates of early transmission media and methodologies, as well as their low throughput, special attention was paid to the development of c...