Segmenting LANs with Repeaters

Segmenting LANs with Repeaters

Legacy Ethernet systems such as 10Base5, 10Base2, and 10BaseT have distance limitations for segments as described in Chapter 1, "Desktop Technologies." Whenever you desire to extend the distance, you can use an internetworking device like a repeater. Repeaters operate at Layer 1 of the OSI model and appear as an extension to the cable segment. Workstations have no knowledge of the presence of a repeater which is completely transparent to the attached devices. A repeater attaches wire segments together as shown in Figure 2-2.

Figure 2-2. Interconnecting LAN Segments with a Repeater

Repeaters regenerate the signal from one wire on to the other. When Station 1 transmits to Station 2, the frame also appears on Wire B, even though the source and destination device coexist on Wire A. Repeaters are unintelligent devices and have no insight to the data content. They blindly perform their responsibility of forwarding signals from one wire to all other wires. If the frame contains errors, the repeater forwards it. If the frame violates the minimum or maximum frame sizes specified by Ethernet, the repeater forwards it. If a collision occurs on Wire A, Wire B also sees it. Repeaters truly act like an extension of the cable.

Although Figure 2-2 shows the interconnection of two segments, repeaters can have many ports to attach multiple segments as shown in Figure 2-3.

Figure 2-3. A Multiport Repeater

A 10BaseT network is comprised of hubs and twisted-pair cables to interconnect workstations. Hubs are multiport repeaters and forward signals from one interface to all other interfaces. As in Figure 2-2, all stations attached to the hub in Figure 2-3 see all traffic, both the good and the bad.

Repeaters perform several duties associated with signal propagation. For example, repeaters regenerate and retime the signal and create a new preamble. Preamble bits precede the frame destination MAC address and help receivers to synchronize. The 8-byte preamble has an alternating binary 1010 pattern except for the last byte. The last byte of the preamble, which ends in a binary pattern of 10101011, is called the start of frame delimiter (SFD). The last two bits indicate to the receiver that data follows. Repeaters strip all eight preamble bytes from the incoming frame, then generate and prepend a new preamble on the frame before transmission through the outbound interface.

Repeaters also ensure that collisions are signaled on all ports. If Stations 1 and 2 in Figure 2-2 participate in a collision, the collision is enforced through the repeater so that the stations on Wire B also know of the collision. Stations on Wire B must wait for the collision to clear before transmitting. If Stations 3 and 4 do not know of the collision, they might attempt a transmission during Station 1 and 2's collision event. They become additional participants in the collision.

Limitations exist in a repeater-based network. They arise from different causes and must be considered when extending a network with repeaters. The limitations include the following:

• Shared bandwidth between devices

• Specification constraints on the number of stations per segment

• End-to-end distance capability

Shared Bandwidth

A repeater extends not just the distance of the cable, but it also extends the collision domain. Collisions on one segment affect stations on another repeater-connected segment. Collisions extend through a repeater and consume bandwidth on all interconnected segments. Another side effect of a collision domain is the propagation of frames through the network. If the network uses shared network technology, all stations in the repeater-based network share the bandwidth. This is true whether the source frame is unicast, multicast, or broadcast. All stations see all frames. Adding more stations to the repeater network potentially divides the bandwidth even further. Legacy Ethernet systems have a shared 10 Mbps bandwidth. The stations take turns using the bandwidth. As the number of transmitting workstations increases, the amount of available bandwidth decreases.

Note

Bandwidth is actually divided by the number of transmitting stations. Simply attaching a station does not consume bandwidth until the device transmits. As a theoretical extreme, a network can be constructed of 1,000 devices with only one device transmitting and the other 999 only listening. In this case, the bandwidth is dedicated to the single transmitting station by virtue of the fact that no other device is transmitting. Therefore, the transmitter never experiences collisions and can transmit whenever it desires at full media rates.

It behooves the network administrator to determine bandwidth requirements for user applications and to compare them against the theoretical bandwidth available in the network, as well as actual bandwidth available. Use a network analyzer to measure the average and peak bandwidth consumed by the applications. This helps to determine by how much you need to increase the network's capacity to support the applications.

Number of Stations per Segment

Further, Ethernet imposes limits on how many workstations can attach to a cable. These constraints arise from electrical considerations. As the number of transceivers attached to a cable increases, the cable impedance changes and creates electrical reflections in the system. If the impedance changes too much, the collision detection process fails. Limits for legacy systems, for example, include no more than 100 attached devices per segment for a 10Base5 network. A 10Base2 system cannot exceed 30 stations. Repeaters cannot increase the number of stations supported per segment. The limitation is inherent in the bus architectures of 10Base2 and 10Base5 networks.

End-to-End Distance

Another limitation on extending networks with repeaters focuses on distance. An Ethernet link can extend only so far before the media slotTime specified by Ethernet standards is violated. As described in Chapter 1, the slotTime is a function of the network data rate. A 10 Mbps network such as 10BaseT has a slotTime of 51.2 microseconds. A 100 Mbps network slotTime is one tenth that of 10BaseT. The calculated network extent takes into account the slotTime size, latency through various media such as copper and fiber, and the number of repeaters in a network. In a 10 Mbps Ethernet, the number of repeaters in a network must follow the 5/3/1 rule illustrated in Figure 2-4. This rule states that up to five segments can be interconnected with repeaters. But only three of the segments can have devices attached. The other two segments interconnect segments and only allow repeaters to attach at the ends. When following the 5/3/1 rule, an administrator creates one collision domain. A collision in the network propagates through all repeaters to all other segments.

Figure 2-4. Interconnecting with the 5/3/1 Rule

Repeaters, when correctly used, extend the collision domain by interconnecting segments at OSI Layer 1. Any transmission in the collision domain propagates to all other stations in the network. A network administrator must, however, take into account the 5/3/1 rule. If the network needs to extend beyond these limits, other internetworking device types must be used. For example, the administrator could use a bridge or a router.

Repeaters extend the bounds of broadcast and collision domains, but only to the extent allowed by media repeater rules. The maximum geographical extent, constrained by the media slotTime value, defines the collision domain extent. If you extend the collision domain beyond the bounds defined by the media, the network cannot function correctly. In the case of Ethernet, it experiences late collisions if the network extends too far. Late collision events occur whenever a station experiences a collision outside of the 51.2 ms slotTime.

Figure 2-5 illustrates the boundaries of a collision domain defined by the media slotTime. All segments connected together by repeaters belong to the same collision domain. Figure 2-5 also illustrates the boundaries of a broadcast domain in a repeater-based network. Broadcast domains define the extent that a broadcast propagates throughout a network.

Figure 2-5. Broadcast and Collision Domains in a Repeater Network

To demonstrate a collision domain, consider IP's Address Resolution Protocol (ARP) process as in Figure 2-6 when IP Station 1 desires to communicate with Station 2. The stations must belong to the same subnetwork as there is no router in the network. Station 1 first ARPs the destination to determine the destination's MAC address. The ARP frame is a broadcast that traverses the entire segment and transparently passes through all repeaters in the network. All stations receive the broadcast and therefore belong to the same broadcast domain. Station 2 sends a unicast reply to Station 1. All stations receive the reply because they all belong to the same collision domain (although it is handled by the NIC hardware as discussed in Chapter 1).

Figure 2-6. ARP Operation in a Repeater Network