Switch/Router Architectures
Description
This book delves into the inner workings of router and switch design in a comprehensive manner that is accessible to a broad audience. It begins by describing the role of switch/routers in a network, then moves on to the functional composition of a switch/router. A comparison of centralized versus distributed design of the architecture is also presented. The author discusses use of bus versus shared-memory for communication within a design, and also covers Quality of Service (QoS) mechanisms and configuration tools.
Written in a simple style and language to allow readers to easily understand and appreciate the material presented, Switch/Router Architectures: Shared-Bus and Shared-Memory Based Systems discusses the design of multilayer switches--starting with the basic concepts and on to the basic architectures. It describes the evolution of multilayer switch designs and highlights the major performance issues affecting each design. It addresses the need to build faster multilayer switches and examines the architectural constraints imposed by the various multilayer switch designs. The book also discusses design issues including performance, implementation complexity, and scalability to higher speeds. This resource also:
* Summarizes principles of operation and explores the most common installed routers
* Covers the design of example architectures (shared bus and memory based architectures), starting from early software based designs
* Provides case studies to enhance reader comprehension
Switch/Router Architectures: Shared-Bus and Shared-Memory Based Systems is an excellent guide for advanced undergraduate and graduate level students, as well for engineers and researchers working in the field.
More details
Other editions
Additional editions
Person
JAMES AWEYA, PHD, is a Chief Research Scientist at Etisalat British Telecom Innovation Center (EBTIC) in Abu Dhabi, UAE. He has been granted 63 US patents and has published over 54 journal papers, 39 conference papers, and 43 Nortel technical reports. Dr. Aweya is a Senior Member of IEEE.
Content
About the Author vii
Preface ix
1 Introduction to Switch/Router Architectures 1
2 Understanding Shared-Bus and Shared-Memory Switch Fabrics 17
3 Shared-Bus and Shared-Memory-Based Switch/Router Architectures 43
4 Software Requirements for Switch/Routers 61
5 Architectures with Bus-Based Switch Fabrics: Case Study-DECNIS 500/600 Multiprotocol Bridge/Router 87
6 Architectures with Bus-Based Switch Fabrics: Case Study-Fore Systems Powerhub Multilayer Switches 111
7 Architectures with Bus-Based Switch Fabrics: Case Study-Cisco Catalyst 6000 Series Switches 129
8 Architectures with Shared-Memory-Based Switch Fabrics: Case Study-Cisco Catalyst 3550 Series Switches 151
9 Architectures with Bus-Based Switch Fabrics: Case Study-Cisco Catalyst 6500 Series Switches with Supervisor Engine 32 171
10 Architectures with Shared-Memory-Based Switch Fabrics: Case Study-Cisco Catalyst 8500 CSR Series 191
11 Quality of Service Mechanisms in the Switch/Routers 213
12 Quality of Service Configuration Tools in Switch/Routers 227
13 Case Study: Quality of Service Processing in the Cisco Catalyst 6000 and 6500 Series Switches 249
Appendix A: Ethernet
Appendix B: IPv4 Packet
References
Index
1
Introduction to Switch/Router Architectures
1.1 Introducing the Multilayer Switch
The term multilayer switch (or equivalently switch/router) in this book refers to a networking device that performs both Open Systems Interconnection (OSI) network reference model Layer 2 and Layer 3 forwarding of packets (Figure 1.1). The Layer 3 forwarding functions are typically based on the Internet Protocol (IP), while the Layer 2 functions are based on Ethernet. The Layer 2 forwarding function is responsible for forwarding packets (Ethernet frames) within a Layer 2 broadcast domain or Virtual Local Area Network (VLAN). The Layer 3 forwarding function is responsible for forwarding an IP packet from one subnetwork, network or VLAN to an another subnetwork, network, or VLAN.
Figure 1.1 Layer 2 forwarding versus Layer 3 forwarding.
The IP subnetwork could be created based on well-known IP subnetworking rules and guidelines or as a VLAN. A VLAN is a logical group of devices that can span one or more physically separate network segments that are configured to intercommunicate as if they were connected to one physical Layer 2 broadcast domain. Even though the devices may be located on a number of different physical or geographically separate network segments, the devices can intercommunicate as if they are all connected to one physical broadcast domain.
For the Layer 3 forwarding functions to work, the routing functions in the multilayer switch learn about other networks, paths to destination networks and destinations, through dynamic IP routing protocols or via static/manual configuration information provided by a network administrator. The dynamic IP routing protocols - RIP (Routing Information Protocol), OSPF (Open Shortest Path First) Protocol, IS-IS (Intermediate System-to-Intermediate System) Protocol, BGP (Border Gateway Protocol) - allow routers and switch/routers to communicate and distribute network topology information between themselves and provide updates when the network topology changes occur. The routers and switch/routers via the routing protocols learn about the network topology to try to select the best loop-free path on which to forward a packet from its source to its destination IP address.
1.1.1 Control and Data Planes in the Multilayer Switch
The Layer 3 and Layer 2 forwarding functions can each be split into subfunctions - the control plane and data (or forwarding) plane functions (Figure 1.2). Comprehensive discussion of the basic architectures of routers is given in [AWEYA2000] and [AWEYA2001]. The Layer 2 functions in an Ethernet switch and switch/router involve relatively very simple control and data plane operations.
Figure 1.2 Control and data planes in a multilayer switch.
The data plane operations in Layer 2 switches involve MAC address learning (to discover the ports on which new addresses are located), frame flooding (for frames with unknown addresses), frame filtering, and frame forwarding (using a MAC address table showing MAC address to port mappings). The corresponding control plane operations in the Layer 2 devices involve running network loop prevention protocols such as the various variants of the Spanning Tree Protocol (STP), link aggregation-related protocols, device management and configuration tools, and so on.
Even though the Layer 2 functions can be split into two planes of control and data operations, this separation (of control plane and data plane) is usually applied to the Layer 3 functions performed by routers and switch/routers. In a router or switch/router, the entity that performs the control plane operations is referred to as the routing engine, route processor, or control engine (Figure 1.3).
Figure 1.3 Control and forwarding engines in multilayer switches.
The entity that performs the data (or forwarding) plane operations is referred to as the forwarding engine or forwarding processor. By separating the control plane operations from the packet forwarding operations, a designer can effectively identify processing bottlenecks in the device. This knowledge allows the designer to develop and/or use specialized software or hardware components and processors to eliminate these bottlenecks.
1.1.2 Control Engine
Control plane operations in the router or switch/router are performed by the routing engine or route processor, which runs the operating system software that has modules that include the routing protocols, system monitoring functions, system configuration and management tools and interfaces, network traffic engineering functions, traffic management policy tools, and so on.
The control engine runs the routing protocols that maintain the routing tables from which the Layer 3 forwarding table is generated to be used by the Layer 3 forwarding engine in the router or switch/router (Figure 1.4). In addition to running other protocols such as PIM (Protocol Independent Multicast), IGMP (Internet Group Management Protocol), ICMP (Internet Control Messaging Protocol), ARP (Address Resolution Protocol), BFD (Bidirectional Forwarding Detection), and LACP (Link Aggregation Control Protocol), the control engine is responsible for maintaining sessions and exchanging protocol information with other router or network devices.
Figure 1.4 Routing protocols and routing table in the control engine.
The control engine typically is the module that provides the control and monitoring functions for the entire router or switch/router, including controlling system power supplies, monitoring and controlling system temperature (via cooling fans), and monitoring system status (power supplies, cooling fans, line cards, ports and interfaces, primary/secondary router processors, primary/secondary forwarding engines, etc.). The routing engine also controls the router or switch/router network management interfaces, controls some chassis components (e.g., hot-swap or OIR (online insertion and removal) status of components on the backplane), and provides the interfaces for system management and user access to the device.
In high-performance platforms, more than one routing engine can be supported in the router switch/router (Figure 1.5). If two routing engines are installed, one typically functions as the primary (or master) and the other as the secondary (or backup). In this redundant routing engine configuration, if the primary routing engine fails or is removed (for maintenance/repairs) and the secondary routing engine is configured appropriately, the latter takes over as the master routing engine.
Figure 1.5 Multilayer switch with primary and secondary routing engines.
Typically, a router or switch/router supports a set of management ports (e.g., serial port, 10/100?Mb/s Ethernet ports). These ports, generally located on the routing engine module, connect the routing engine to one or more external devices (e.g., terminal, computer) on which a network administrator can issue commands from a command-line interface (CLI) to configure and manage the device. The routing engine could support one or more USB ports that can accept a USB memory device that allows for the loading of the operating system and other system software.
In our discussion in this book, we consider the management plane as part of the control plane - not a separate plane in its own right (Figure 1.6). The management plane is considered a subplane that supports the functions used to manage the router or switch/router via some connections to external management devices (a terminal or computer). Examples of protocols supported in the management plane include Simple Network Management Protocol (SNMP), Telnet, File Transfer Protocol (FTP), Secure FTP, and Secure Shell (SSH). These management protocols allow configuring, managing, and monitoring the device as well as CLI access to the device.
Figure 1.6 Control plane versus management plane.
A console port (which is an EIA/TIA-232 asynchronous serial port) could allow the connection of the routing engine to a device with a serial interface (terminal, modem, computer, etc.) through a serial cable with an RJ-45 connector (Figure 1.7). An AUX (or auxiliary) port could allow the connection of the routing engine (through a serial cable with an RJ-45 connector) to a computer, modem, or other auxiliary device. Furthermore, a 10/100?Mb/s Ethernet interface could connect the routing engine to a management LAN (or a device that has an Ethernet connection) for out-of-band management of the router or switch/router.
Figure 1.7 Management ports.
The routing table (also called the Routing Information Base (RIB)) maintains information about the network topology around the router or switch/router and is constructed and maintained from information obtained from the dynamic routing protocols, and static routes configured by the network administrator. The routing table contains a list of routes to particular IP network destinations (or IP address prefixes). Each route is associated with a metric that is a "distance" measure used by a routing protocol in performing the best path computation to a destination.
The best path to a destination is determined by a routing protocol based on metric (quantitative value) it uses to "measure" the distance it takes to reach a destination. Different routing protocols use different metrics to measure the...
