[4 pages — 3056 words]

INTRODUCTION

The network diameter of an Ethernet network can be increased using repeaters as long as the network diameter does not exceed the collision domain of Ethernet. All Ethernet nodes must be able to recognize the occurrence of a collision regardless of the physical location of the nodes since the detection of collisions is fundamental in the manner Ethernet arbitrates media access. In this lesson, the concept of switching will be introduced as an alternative to the deployment of repeaters. Switches can not only increase the overall network diameter, but will improve the performance of Ethernet networks as well.

CLASSIFYING DEVICES

Although from the outside a switching hub looks very much like a repeating hub, they are from different classes of equipment. If you study the OSI Communications Model, you will notice seven distinct layers corresponding to different communication services.

At the lowest layer you have the physical layer which is concerned with the actual signals on the medium that represent data. These signals are called symbols and repeaters or repeating hubs receive these symbols and recondition them when extending networks. References such as 10BASE5 and 10BASE-T are physical layer standards.

Above the physical layer is the data link layer which handles the actual transmission and reception of frames sent and received over the physical layer. Issues such as station addressing (MAC or medium access control), framing of the data and error detection are handled by the data link layer. The IEEE 802.3 standard is basically a data link standard although references to physical layer standards are included as well. Bridges operate at the data link layer. A bridge and a switch are one in the same.

Above the data link layer is the network layer which addresses the issues of transferring data, not over just one data link, but over multiple data links. This is classified as internetworking with the Internet Protocol (IP) being the most popular internetworking protocol. Routers are used to direct traffic between multiple data links and the transmission units are called packets. Switches do not commonly operate at this layer but there is such a thing as a layer 3 switch. This is actually a router with some switching functionality.

In terms of hardware, the next layer of interest is the application layer seven. This is where the gateways reside when it is necessary to interconnect dissimilar networks and dissimilar protocols. Gateways are aware of the actual application being run while all other devices such as repeaters, bridges and routers are not. We will concentrate only on one class of device called the bridge.

BRIDGE (SWITCH) CONSTRUCTION

A switch is a bridge and the terms will be used interchangeably. The original bridges were two port devices interconnecting two similar data links to form one larger data link. If this can be accomplished without disruption, the bridge is considered a transparent bridge since communication within a data link or between data links appears the same. You may think that we are describing a router but we are not. A router would consider each data link as an actual network with a corresponding network address. A bridge considers each individual data link as part of one larger data link or one network. The concept of network addressing is not used and individual station addresses (MAC addresses) are not duplicated among the various data links. Unlike the traditional bridge with two ports, the switch has several ports and is usually referred to as a switching hub or just a switch.

Unlike a repeating hub, a switch has basically the same Ethernet interface on each of its ports as found on an Ethernet host adapter. That is because each port must function just like another Ethernet device. It must be able to receive and decode Ethernet frames and test for frame integrity as well as assemble and transmit Ethernet frames. However, each port does not necessarily require its own MAC address as would be required by an Ethernet host adapter. Each switch port functions in promiscuous mode by receiving all frames on its port independent of destination MAC address. During transmissions, the Ethernet port masquerades as the originating device by assuming its source address. Therefore, each port on the switching hub does not require its own MAC address unless bridge addressing is required (the spanning tree algorithm requires bridge addressing).

By having an Ethernet interface on each port, the Ethernet collision domain terminates at the switch port. With a repeating hub, the complete hub is part of the collision domain. By having a switch, the effective network diameter can double with the addition of one switch. This is because the network can be broken into two distinct data links. This is one benefit of switches. The effective network diameter can be increased with the addition of switches. This is especially important at 100 Mbps since the collision domain is only 205 meters wide for copper-based systems.

Another difference between a repeating hub and a switch is that the repeating hub must operate at only one speed- either 10 Mbps or 100 Mbps. A switching hub can have multi-speed ports which can adjust to the capabilities of the device attached to its port. This is called auto-negotiation and different speeds on different ports are allowed. Some switches have fixed low speed ports (10 Mbps) and one or more high-speed ports (100 Mbps) for connection to servers where most of the traffic will be experienced.

By terminating collision domains at each of its ports, a switch effectively segments the network into separate collision domains. If only one device is attached to a switch port (an Ethernet host adapter or another switch port), this is called microsegmentation. Under these circumstances full-duplex operation is possible yielding no collisions. However, if a shared Ethernet collision domain is present on a switch port (multiple host adapters and a repeating hub), only half-duplex operation is allowed and the switch port must conform to Ethernet's medium arbitration rules.

(No part of this article may be reproduced without the written consent of the Industrial Ethernet University.)