CET1173C Lecture #6 - Networking Devices

Materials:
Complete working system with Windows 98
Network Interface Card
Straight-through UTP patch cable
Hub or Switch
Objectives:
The student should become familiar with:
The general concepts and functionality of the various networking devices,
The OSI model layers of each device,
The installation and configuration of networking devices,
The effect of each device on the network,
The effect of changing from one device to another.
Competency:
The student will become familiar with the general nature and functionality of the various types of networking devices including network interface cards, hubs, repeaters, bridges, switches, routers, brouters, and gateways. The student will understand the role each plays on a network, the OSI model layer in which they operate, and the basic function of each device. The student will be able to determine which device will solve a given network design goal and understand the effect of introducing each device to a network.

Lecture

  1. This module will cover all of the basic networking connectivity devices that make it possible for microcomputers to exchange information and therefore participate in a network. The first such device with which the student is already familiar is the network interface card or NIC. There are two major subdivisions of NICs: internal and external. External NICs are generally speaking easier to physically install since they rarely require the PC's case to be removed, while internal NICs require more planning and of course physical installation. The planning aspect for the installation of a NIC involves:

    • Choosing the network type - 10BaseX (many of these NICs have two or more connectors allowing them to interface into either 10BaseT, 10Base2, or 10Base5 networks), 100BaseTX, 16Mbps IBM Token Ring, etc
    • Choosing the expansion bus type - ISA, PCI, PC-Card, Cardbus, USB, etc.
    • Ensuring that the system has an available expansion slot
    • The NIC can use flexible resource assignments - hardware compatibility with the current configuration. This is especially critical with ISA NIC's.
    • The NIC has drivers for the operating system version - software compatibility with the operating system. Most devices are compatible with the generation of operating system that was popular during their manufacture and useful operational lifespan. It would therefore not be a surprise to learn that an ISA 10BaseT NIC does not have Windows 2000 drivers available from the manufacturer. Likewise, it should come as no surprise that a new PCI 100BaseTX adapter come with the drivers for only Windows 2000 and Windows XP and will not work on a Windows 98 PC.
    • The NIC's manufacturer has comprehensive on line support - This should be a consideration for all PC and networking hardware purchases. If the manufacturer can at least provide a driver download, FAQ (Frequently Asked Questions), and firmware revisions (if they exist) online then the NIC will more than likely be much easier to support.
    • The NIC's functional features - NIC's have a wide range of features, many of which can dramatically improve performance.

  2. In upgrading older systems, the existing NIC's might be 10BaseT ISA bus Ethernet adapters. But the systems have an available PCI expansion slot and are therefore capable of being upgraded with a modern 100BaseT PCI bus Ethernet adapter. Another example situation may exist on a server in which the available expansion slots are all PCI-X slots: 64-bit @ 100Mhz PCI. This is a very high performance bus and the average retailer will not carry any NICs that would work on this expansion bus. The server's documentation will indicate if the expansion bus is backwards compatible to accept regular PCI-32bit/33Mhz cards. Laptops accept two major bus types of NICs: PC Card (the technology formerly known as PCMCIA) and Cardbus. PC Card is a 16-bit technology that runs roughly the same speed as ISA while Cardbus is a 32-bit technology that runs at 33Mhz and has roughly the same performance as PCI. A laptop that accepts Cardbus cards could be designed backwards compatible, but in general older laptops accept only PC Card peripherals and newer laptops only accept Cardbus peripherals. Slots that are not compatible with the other type of technology are of a slightly different form factor so if the card is not sliding in easily do not force it: it is the wrong type. The type of bus on the laptop must be ascertained prior to considering NICs for the computer.

  3. NIC's can offer such features as:

    • Automatic speed detection - Since 10BaseT and 100BaseT Ethernet can use identical connectors and cabling, many NIC's are capable of joining either type of network. This feature allows these cards to automatically detect which network they are attached to without having to set it or configure it manually.
    • Diagnostic LED's - A good NIC should have at least a "Link" LED that indicates that the NIC has detected the network's carrier signal. This makes physical connectivity troubleshooting much easier. Dual speed NIC's should also indicate the network speed that they are running at or have autodetected.
    • Autosense - Usually packaged with the drivers allows the NIC to determine what Ethernet frame type is being used on the network and then it will automatically use it also. This makes mismatched frame-type issues almost non-existent.
    • Onboard CPU - allows the NIC to perform frame creation and capture entirely on the NIC rather than have to rely on the system memory and processor. Greatly improves NIC performance and the host system performance.
    • RAM memory buffering - allows the transfer of data to and from the NIC to be offloaded to the NIC's CPU. This is usually integrated into any NIC that has an onboard CPU.
    • "Look Ahead" Processing - allows the NIC's CPU to begin assembly or disassembly of a packet as the information starts to arrive. This is usually integrated into any NIC that has an onboard CPU.
    • Network Management Capabilities - Usually in the driver bundle. Allows the NIC to participate in a global network management system using SNMP.
    • Power Management Capabilities - Allows the NIC to participate in the system's ACPI functions to be able to power down when not in use and also to awaken the system if data begins to arrive: a feature called "Wake-On-LAN".
    • Onboard flashBIOS- Allows the NIC to support the ability to boot the system across the network rather than from a local drive. This was very popular for implementing "disk-less workstations" and is still popular for the purposes of booting from a server in order to run diagnostics, virus scans or restore from backup utilities. Another use of the onboard flashBIOS would be to provide the working program code for an onboard CPU and revisions to this code could provide greater reliability, compatibility and performance.

  4. Older 10Base5 NIC's may not have the actual network transceiver circuitry that actually makes the transmission and detects and receives packets on the card itself. The Ethernet transceiver for such a NIC is a separate device that attaches to the RG-8 cabling using a vampire tap. Then a drop cable is attached to this transceiver and this is what attaches to the NIC. This is the type of NIC that uses an AUI - Attachment Unit Interface connector which on the NIC has a female DB15 connector.

  5. Since a NIC under normal circumstances does have the transceiver circuitry and is also equipped with a built in MAC address a NIC functions at both the OSI Layer 1 - Physical Layer and at the OSI Layer 2 - Data Link layer. When given the choice between calling a NIC an OSI L1 device or an OSI L2 device, then it should be called an OSI L2 device - the highest OSI model layer in which it functions.

  6. Repeaters are network devices that increase the distance that a network can span by regenerating a signal that arrives in one port and transmitting out the other port. A 10Base2 physical segment is limited to a total length of 185 meters because a signal from one end to the other will experience too much attenuation such that it will no longer consist of clean well defined waves beyond this distance. A repeater on the end of this segment will receive the weakened signals and retransmit them as clear pure waves again on the other port. This allows two 185M segments to be joined by a repeater. In fact, repeaters could be placed at the ends of several segments in a daisy chain which prompted the development of the 5-4-3 rule for Ethernet networks so that the total length and node population of the network would not exceed the ability of devices on opposite ends from being able to detect collisions with each other.

  7. Repeaters only pay attention to the individual waves of the signal transmission and faithfully regenerate these waves on the other port. They do not interpret any part of the Ethernet frames and are unaware of MAC addresses and do not themselves have a MAC address. Therefore repeaters are strictly OSI Layer 1 - Physical Layer Devices.

  8. Hubs or concentrators as they used to be called in the olden days, are essentially nothing more than multiport repeaters. They provide an individual connectivity port to each network node using UTP or STP cable rather than coaxial cabling on the network. Because network grade UTP generally consists of 4 twisted pairs of wires to achieve resistance to noise it is fairly difficult to splice into the cable in the manner that is done with the coaxial cable physical bus topologies. Instead the eight conductors are attached to an eight pin connector called the RJ-45 (Registered Jack) connector. The cable can attach to a single port at either end. Because of this restriction another method would be needed in order to facilitate the assembly of a network containing multiple nodes using Ethernet which is a logical bus topology in which all nodes must be able to listen for a clear line on which to transmit, and all must be able to transmit at any time to any other physical node on the segment.

  9. Hubs provide a centralized connectivity point to which all nodes on the segment can be attached. The hubs circuitry is a matrix of circuits designed to cross connect all of the ports to each other so that the Ethernet logical bus is preserved. All nodes are effectively on the same circuit through the hub and can all listen to the line that leads to their NIC and determine that the segment is clear and begin sending traffic. The hub like the repeater only facilitates the passage of all of these signals through it including a collision and jamming signals. Because of this and the fact that the hub does not inspect the packets information at any level it is also considered an OSI Layer 1 - Physical layer device.

  10. There are four types of hubs and these definitions get very blurry from one author to the next. These definitions will attempt to conform to the general concensus of the maajority of authors.

    • Passive - A passive hub does not regenerate the signals that arrive at the ports and retransmit new fresh signals out of all of the other ports. Instead it simply ties all of the ports wiring together so that a signal arriving on one port will pass through and out all of the other ports with the signal unchanged and unprocessed by the hub. Because of this the total distance of the 10BaseT segment is not only from the node to the hub, but includes the distance of the cable to the other node as well. Therefore the cable from node "A" to the hub and the cable from node "B" to the hub must not exceed a total length of 100 meters. This means that if node "A" is 75 meters away, then node "B" must be no more 25 meters away. In fact no other node can exceed 25 meters or it will be unreachable by node "A". Passive hubs are not even therefore technically speaking multport repeaters. They were very popular in the early 10BaseT days because they were much cheaper than active hubs. They should be quite rare by now and I don't believe that there are any 100BaseTX passive hubs.
    • Active - An active hub is a true multiport repeater in that the signal that arrives on one port will be regenerated by the onboard circuitry and retransmitted fresh out of all of the other ports. This allows the maximum segment length of 10BaseT to be considered from the node to the active hub only. This means that all nodes can be 100 Meters away from the device and effectively reside on their own segments with themselves at one end and the active hub repeater at the other end.
    • Managed hub - This device allows an interface through management software running on a workstation on the network so that it can be monitored for netork traffic and statistics as well as be configured. Otherwise, they are usually functionally equivalent to active hubs.
    • Intelligent hub - A specialized device for use on 100BaseVG-AnyLAN networks. These networks were a proprietary technology developed in the Ethernet speed race of the early '90's. 100BaseVG can achieve 100Mbps throughput using existing CAT3 UTP cable which 100BaseTX cannot do. It does this by using three of the four pairs of wires to transmit at 33Mbps in one direction at a time - half duplex only. However, 100BaseVG is not a form of Ethernet at all and depends on the hub as an integral participant in the function of the network in which the device does inspect the contents of the packets and it will forward packets to their destination using a demand priority logical network topology that is neither a bus, nor a ring, but closer to a star or mesh. This technology fell out of favor and should be very hard to find these days and is defined by the IEEE 802.12 specification.

  11. Hubs can feature any of the following:

    • Port LED's - A good hub should at least have a "Link" LED for each port indicating whether an active node is attached to the end of the cable that is attached to the port. Good hubs also feature "Traffic" LED's that indicate actual transmissions as they occur from the attached node. Good hubs can also feature a "Collision" LED that indicates which nodes have collided.
    • Uplink Port - The uplink port is internally cross wired so that the transmit pair of wires in the cable attach to the receive wires of all other ports and vice versa. This allows another hub to be attached to this hub from a regular port to this one using a straight-through cable. In this way up to four hubs (the 5-4-3 rule limit) can be daisy-chained without needing cross-over cables to do it. Most hubs uplink ports are a shared port with the nearest normal port. In this case either the nearest normal port or the uplink port can be used but never both at the same time since they are nothing more than two separate RJ-45 connectors to the same port circuitry internally. Some good hubs have autodetecting uplink ports that will switch to internal cross-over wiring if needed, these may or may not be shared ports, this will have to be determined by the documentation.
    • Management Port - Since the hub is not a network node and does not have a MAC address of its own, managed hubs may have a management port to which a "management console" can be attached. The interface may depend on proprietary software and even hardware connectors and cables to a PC's serial port, for example. Managed ports may be able to interface through the management port directly or through the management console to the network management software package using SNMP.
    • Power supply connector - Most small hubs have a small phono jack style DC current connector to an external AC/DC transformer power supply that plugs into the wall outlet. Most enterprise level hubs including rack mount units have an internal AC/DC converting power supply like the PC itself and can accept a standard PC power cord and plug directly to the AC power source.
    • Cooling Fan - Most smaller hubs do not have a cooling fan because they do not have high voltage AC/DC converter circuitry onboard. Most enterprise level hubs that do have high voltage AC/DC converter circuitry onboard should also be equipped with a cooling fan.
    • Power Switch and Power Indicator - Active hubs may have a power switch (good ones) or simply switch themselves on and off with the attachment of the power supply cable (cheap ones). All good active hubs should have a clearly visible power on LED indicator.

  12. Hubs come in a variety of form factors:

    • Standalone - Not physically designed to be interfaced or daisy chained to others but they usually can be anyway. These are usually the small low end work group hubs that are used in SOHO (Small Office/Home Office) networks.
    • Stackable - Physically designed to stack one or more of the same model on top of each other providing their own built in support for the stack. These are usually marketed to medium to large enterprises and can range from as little as eight ports to a more typical 12 to 24 port size.
    • Rackmount - Usually built with the standard 19" width so that they can be mounted to modular enterprise level telecommunications closet racks. These are typically 16 ports to much larger units featuring 48 ports or more.
    • Modular - The modular hub is the most sophisticated hub. It can be standalone, stackable or rackmount but since they are also the most expensive form of hub they are usually rackmount and intended for medium to large enterprise networks. A modular hub can be purchased with any number of regular or special function ports installed and can be easily upgraded with more regular or special function ports at a later time defraying the cost of the device until the network grows large enough to actually need the extra ports.

  13. Bridges are now considered vintage Ethernet network devices, but since there are enough vintage networks still in operation the complete entworking technician should know what they are and how they work. Furthermore, understanding how bridges work will help in the fundamental understanding of how Ethernet switches work also. A bridge's main function is to split a single Ethernet segment in half and they usually only have two ports because of this. A bridge can effectively split a collision domain, but it cannot extend it beyond its normal physical size limitation from end to end. So a bridge cannot be used to connect two maximum sized Ethernet collision domains into one. This is because the bridge must still propogate Ethernet packets which must still be able to collide with each other and be detected by both transmitting nodes. So a bridge can replace a repeater between physical segments and they are usually found perched in the middle or at intervals along older 10Base2 and 10Base5 Ethernet networks. Bridges can dramatically improve performance on heavily populated networks because they do filter Ethernet frames. They generally perform this function automatically by watching the source MAC addresses within the packets as they arrive and are forwarded to the other segment that they divide. They record which segment the source MAC resides on so after all four nodes on this network have made a single Ethernet transmission the bridge already knows which segment each one resides on:



  14. Knowing this, if node "1" transmits a packet with the destination MAC address of node "2" then the bridge will not forward the packet to segment "B" and it will perform this filtering for all packets arriving at either port. Therefore the bridge effectively cuts the segment in half. It would be a poor idea to have all of the workstations on one side and all of the servers on the other side because the majority of communication on that network would have to cross the bridge nullifying its benefit. However, it could be placed between two logical domains with one server and all of its clients on one side and another server and all of its clients on the other. They can all still communicate when necessary but the vast majority of communications in this scenario are limited to the intended devices on the local segment and the bridge can filter them all out and greatly reduce the overall amount of traffic that any node experiences. If a node on segment "A" is receiving a huge file streaming from the server on segment "A" then segment "B" will be experiencing no traffic at all and a node on segment "B" could initiate a session with the server on segment "B" at the same time. This leads to far greater throughput and performance across the network.

  15. Since bridges inspect and understand Ethernet frames looking at the MAC addresses and perform intelligent decisions based on this Data Link layer information they are considered to be OSI Layer 2 - Data Link layer devices. Bridges that begin working automatically by simply collecting MAC address lists off of each port and perform no further filtering techniques than that are refered to as transparent bridges.

  16. In a large mesh or partial mesh network where a packet could conceivably arrive at two different bridges and get forwarded to the otherside where it runs into more bridges could end up being batted back and forth between the bridges in an endless bridging loop that could bring the entire network to a halt. Once this problem was discovered bridges were redesigned to implement a technique called the spanning tree algorithm. This allows bridges to detect each other and cooperate and automatically discover possible circuitous paths that packets might accidentally end up bridge looped within and the bridges will arrive at a concensus on how to forward packets in the network in these situations. The bridges can discover each other or new ones and incorporate themselves into a network scheme rather quickly and do not seem to be a source of noticeable additional traffic.

  17. Another type of intelligent bridge is used on token ring networks and is referred to as a source-route bridge. These bridges also discover each other and determine the best route for packets to take based on the source and destination MACs. However, source-route bridges add additional routing information into the packets adding a small amount of overhead to the network. But since a token ring network only has one sending and one receiving node at any given moment this does not seem to adversely affect traffic like it might on an Ethernet network. The fourth type of intelligent bridge found on networks is a translational bridge. These are able to convert one Layer 2 technology into another. For example, Ethernet frames are completely different from IBM Token Ring frames but each of these Data Link layer frames can carry the same unchanged IP packet. A translational bridge would sit between the Ethernet network and the IBM Token Ring network and pick up an Ethernet frame remove the IP packet from it and rebroadcast it on the IBM Token Ring network within a Token Ring frame and vice versa when necessary. Obviously a translational bridge watches the MAC addresses in the headers of the Token Ring and Ethernet frames and only forwards packets as necessary between the two different networks.

  18. Switches can be thought of as multiport bridges. The devices themselves look like hubs and in fact can replace hubs in 10BaseT and 100BaseTX Ethernet networks. Functionally however, they are not hubs at all. A hub will rebroadcast the packet that is arriving on one port back out all of the other ports simultaneously; thereby simulating the Ethernet bus topology (logical topology of Ethernet is a bus) through the star physical topology. The switch like the bridge watches the Ethernet frames that arrive through each port to get the MAC address of the device attached to that port. Once it has built the table of MACs to physical ports when a packet arrives on any port and the destination MAC address is in the lookup table the switch will rebroadcast the packet only to that destination port and not the others. This effectively isolates the sender and the receiver onto their own privately switched circuit which forms their own collision domain, hence the name. Any other node can begin a transmission and if the packet is not intended for either currently occupied node, the switch will send the packet to that destination node as well.

  19. Switches therefore dramatically improve overall network performance and throughput on the segment and function like transparent bridges in that they automatically assemble their MAC-to-port tables as soon they are they connected and turned on. Switches also dramatically improve physical/data link layer security by preventing packets between two particular nodes from being captured by an arbitrary third node like can be done when using a hub which rebroadcasts all packets to the entire segment. If the network is a large branched star using only hubs then any node on the entire LAN can capture any transmitted packet. With switches which only forward the packet to the destination port, this packet grabbing capability is totally eliminated. Since switches greatly improve performance and security most 10BaseT and 100BaseTX networks have already changed out all of the hubs with switches. Since 10/100Mbps switches currently cost as little as $10/port any netowrk still using hubs should be changed out with switches.

  20. There are two basic types of switching that switches perform. Good switches may allow the type of switching to be changed from one form to the other, but it will most likely be implemented on all ports not selected ports. Cut through switches begin intepreting the frame immediately upon arrival. That is, after the preamble and the start frame delimiter the next piece of information is the destination MAC address. As it arrives the switch will read it, look it up in the MAC-to-port table, and immediately begin sending it out that port before the end of the frame has been read by the switch. Cut through switches are therefore very fast, but they will propagate a bad frame.

  21. Store and forward switches will receive and process the entire frame first. As it receives it it looks up the destination port. Once the entire frame has been received it is capable of performing the frame check sequence calculation to ensure that the frame is a good one. Once this has been determined it then forwards it to the destination port. Store and forward switches are slightly slower and could slightly increase frame latency (the delay between transmission and final receipt by the destination node) on very large networks. However, all bad frames can be filtered out by the switch which will prevent target nodes of the bad frames from having to deal with them. So overall network performance can be better with store and forward switches when the network is faced with any situation that may be causing a stream of bad packets. Under normal situations the cut through switch will provide slightly better performance.

  22. If a node begins sending a frame and the destination node is currently involved in the exchange of a frame, then if the switch cannot store the frame, like a cut through switch, it will transmit a jamming signal back down the originating port to cause a collision to force the sender to wait and then try again. Store and forward switches may automatically jam a sender to an occupied node as well just to keep from ending up having to hold on to more and more frames until it would be forced to drop them (run out of RAM to store them). It is preferable to take advantage of the low level functionality of Ethernet and jam the sender. Remember that these situations will arrise more often because the switch leaves a quiet line open to all nodes that are not currently involved in the exchange of the frame. Whereas with a hub or regular bus topology as the two nodes exchange the packet, the other nodes hear the exchange going on and do not attempt to transmit. Individual nodes on a heavily crowded Ethernet LAN using switches should not surprisingly report more collisions than a network using hubs.

  23. Switches can be used to define VLANs - Virtual LANs. Switches can be instructed to allow a certain group of MACs or ports to be able to establish contact with each other and also prevent all other MACs or ports from being able to contact the group. In this way the switch can impose a complete Ethernet network for the chosen group of ports that excludes connectivity to all other devices attached to the switches. Any number of ports can be placed into one of these groups and any number of groups can be created up to the number of ports available. The switches must be capable of being configured in this manner which most smaller switches can not do. The Virtual LAN begin defined is not limited to the ports of a single switch and can include ports from any switch on the collision domain and is configured using the manufacturer's configuration utilities. In theory, the only way that a member of a VLAN could communicate with nodes not on the VLAN would be through some form of router that is attached to the ports of both VLANs. VLANs are often used to improve throughput and to create related and organized groups of nodes that need to be networked while excluding nodes that normally do not need to be networked to them. Switches work with MAC addresses and entire Ethernet frames and are therefore OSI Layer 2 - Data Link layer devices.

  24. Higher layer switches are also available. A Layer 3 Switch can set up VLANs as well as filter traffic based on IP addresses. These switches are also used for statistics gathering and advanced security functions. Theoretically they are simple routers, but true routers are usually far more sophisticated in their total range of features and capabilities. Routers are usually capable of interfacing different networks like Token Ring and Ethernet, they are capable of interfacing Dial-Up or T-1 Carriers traffic to Ethernet, capable of recognizing and interpreting encapsulated packets, and so forth making them much closer to the definition of a gateway, hence the terminology in IP configuration of setting the default gateway which is usually a router. The trade off in capabilities is compensated for by the ease of installation and configuration of the Layer 3 switch and its performance is much faster because it is specialized mostly to pay attention to source and destination IP's and perhaps CRC fields only. There are also Layer 4 switches which can control, monitor and gather statistics based on Transport layer datagrams as well. Again their functions are limited compared to true router/gateway devices and optimized more for speed than for configurable functions.

  25. Routers are multiport connectivity devices originally designed to separate physical networks and to be able to forward the IP packets from one physical network to another. It is the router then that is the actual physical networking device that facilitates the OSI Layer 3 - Network layer making the concept of configurable logical addressing and logical packets be able to leave the confines of the physical network and be forwarded on into another network. For example, at the extreme end of a 2500 Meter 10Base5 Ethernet network another 500 Meter segment cannot be attached, it exceeds the physical limitation of the network and will literally crash the entire network. Instead a router can be attached and become a node of the network with one of its NICs. It can also attach to another complete 2500 Meter long 10Base5 Ethernet network with another NIC. In this way the router has physically joined each Ethernet network separately with each NIC. The router will need routing tables that contain the destination IP addresses for each network. When an Ethernet frame arrives on the NIC on network #1 with an IP packet with a destination address intended for a device on network #2, then the router will read the IP packet out of the Ethernet frame of network #1 which it discards. It will then generate a brand new Ethernet frame with the destination MAC of the machine with the IP of the destination IP of the IP packet. It will then transmit it to the node on network #2. This is the original intent of the router. Many routers are modular and capable of accepting standard NICs as well as modems, ISDN terminal adapters, T-1 carrier CSU/DSU's and so on. The router is the original OSI layer 3 - Network layer device.

  26. Because of the router's ability to interface completely independent physical networks it must often interpret completely different data link layer protocols as well. Because so much data interpretation is required routers are complete computers with their own operating systems, start up sequence, configuration files and so on. Cisco routers for example have their own operating system known as IOS and it advances with versions like any other operating system. Networking professionals who would be expected to work with the router would be expected to be familiar with their operating system in order to be able to configure them properly. Cisco offers their own proprietary certification program which begins with the CCNA - Cisco Certified Network Associate. This is a recommended certification for any network technician even though it is proprietary mainly because Cisco is estimated to hold over 90% of the router market.

  27. Routers have evolved from their beginning as the basic facilitators of the OSI Network layer into very sophisticated and intelligent devices that can operate at layer 4 and even up to layer 7. As such they perform functions that are more accurately ascribed to a network device called a gateway because of this increased higher OSI layer functionality. Routers can perform such operations as:

  28. Routers depend on the routing tables mentioned earlier and it was also mentioned that the router will need to map the data link layer address with the IP address in order to forward packets from one network to another indicating that the router must be capable of performing ARP functions as well. There are two basic types of router based on the source of the routing table that they use: static and dynamic routing tables. A statice routing table is manually entered into the router one entry at a time. This obviously will be highly ineffective on routers that interconnect large networks especially ones using DHCP where client addresses expire and constantly change. A static routing table would be acceptable in a situation where the router only connects to several other routers acting as the central connectivity point of a very large network.

  29. Dynamic routing tables on the other hand are assembled automatically. A router can simply be plugged in between two networks and begin gathering IP address information and assemble a routing table as each node makes a transmission. However, the router may not be able to determine what to do with a packet that needs to travel through many routers before it can reach the desired destination node. To solve this issue routers must be able to share information with each other about IP network numbers. When a packet arrives and contains a destination node with a foreign netowrk number the router will look the network number up in its routing tables. If it cannot find it, then the router will forward the packet to its own default gateway (the next router down the line) and hope for the best. If the routers are set up in branched meshes or partial meshes destination network numbered packets could get caught in loops until their TTL fields expire. See NET1 Lecture #4 - OSI Layer 3 - Network Layer for details.

  30. There are four major protocols used by routers to exchange routing information so that such problems will not occur: RIP, OSPF, EIGRP, BGP

  31. RIP - Routing Information Protocol for IP (and there is a RIP for IPX). RIP is the oldest routing protocol and depends on routers broadcasting their current routing tables to all other routers every 30 seconds by default. This creates the kind of traffic that routers are responsible for stopping (fairly useless broadcasts). RIP also considers only the hops required to reach the destination. Because of this set interval for routing table update broadcasts on large networks it could take several minutes for a node or network address to be updated throughout the network and become accessible again. Therefore a RIP network is said to have a poor convergence time. RIP networks are also not as secure as others since security is not built into the protocol and sensitive information about hidden network numbers could be exposed in the router broadcasts. The benefit for using RIP (which is still used on many routers) is that RIP based networks are highly robust, it is slow, chatty, and insecure, but it does work very well. RIP networks are also fairly fast except for their convergence time so if network numbers rarely change and the routers can be configured to perform RIP broadcasts every 5 minutes for example, then it could be a fairly effective private network routing protocol.

  32. OSPF - Open Shortest Path First for IP. This one was developed specifically for IP networks and has significant improvements to make up for the weaknesses of RIP. In fact OSPF can coexist on the same network as RIP. OSPF does not only use the hop count to determine the best path to the destination node. OSPF routers keep routing table databases of information for all other known routers and will track and update each other only as needed so there is no set interval for broadcasts. Also an OSPF update will only contain the changes to the existing routing information and it will not be a potentially huge broadcast of the routers entire collection of routing tables. OSPF routers can therefore detect a bogged down link and calculate a new route to forward packets through until the traffic jam subsides. Because of this and the fact that OSPF routers broadcasts are much smaller and more specific OSPF networks enjoy a very fast convergence and in general enjoy a far superior overall performance because the routers try to maximize bandwidth availability amongst themselves. So Router #1 might be a way to forward packets to the destination through only two hops, but if it gets bogged down and transmits an update to the others about this, they will find an idle router that can get aroung the problem even though the packets must travel through three hops to get to the destination. OSPF therefore sidesteps the log jam and delivers the packets quickly and maximizes available bandwidth by finding another clear route to send the packets through.

  33. EIGRP - Enhanced Interior Gateway Routing Protocol for IP, IPX and AppleTalk. This is the routing protocol developed by Cisco systems fro use in their routers. EIGRP is the property of Cisco and can only be found on their routers. EIGRP is much more processor intensive which is why Cisco routers in particular are full computers with their own operating system because it can support multiple protocols. EIGRP like OSPF is a routing protocol that features rapid and concise dynamic updates as well as very fast convergence times. The protocol is also designed around the ability of the routers to automatically detect and compensate for bogged down as well as broken links.

  34. BGP - Border Gateway Protocol for IP. BGP has been developed for use by the Internet backbone routers. Due to the rapid explosion of growth of the Internet not only in the total number of nodes but also the total number of networks BGP was developed so that the Internet's backbone routers would be able to manage enormous amounts of traffic between hundreds of thousands of networks across dozens of hops and is by far the most complex and sophisticated routing protocol. BGP generally runs only on the very large, powerful and expensive Internet backbone routers and can interface with all other IP routing protocols seamlessly. Because the Internet is a TCP/IP only network BGP only supports IP and the IP versions of the other routing protocols.

  35. Brouters and routing switches are hybrid devices. A brouter or a routing switch are essentially the same device. A brouter is a router working within OSI layer 3 and can separate physical networks just like a normal router. When a non-routable packet arrives however, the brouter's bridge capabilities(or switch, same functionality) will attempt to bridge the packet to its destination. Obviously the only way that the bridging function will work is if the two physical networks are similar since the bridge will try to forward the data link layer packet to the target node's segment. Brouters are needed on complex networks that use non-routable protocols like NetBEUI. But since the majority of these networks are being converted to use pure IP the market for brouters is dwindling. Routing switches are basically the same device as mentioned above called the layer 3 switch.

  36. Gateways by definition are any network connectivity device or node that can interface totally different networks at any OSI model layer. Because of this gateways require quite a large amount of processing power since they perform sophisticated processing of packets and conversion to other networking formats. Gateways are not solely network hardware devices because of their level of sophistication they a gateway is a combination of hardware and software. Some powerful routers perform the functions of a true gateway such as the ability to interface Ethernet and Token Ring networks. Servers and mainframes can also be set up to perform gateway networking functions.

  37. Gateways are potentially the slowest connectivity device and should only be used when no other solution will suffice. If IP and Layer 4 switching will handle the needs then routers should be used. Some examples of application layer gateways include:

  38. There are aside from the above mentioned network connectivity devices several other dedicated devices that range from quite common to quite rare. The most common of the "other devices" is the wireless access point. These are usually multifunction devices whose functionality is similar to a switch or hub, but not completely the same due to the nature of the medium (radio transmissions through the air). WAPs are covered fully in the WLAN Technologies module.

  39. Another significant networking device that might be encountered is the media converter. This can be a standalone device that accepts the input from one type of physical/datalink type of network on one port and outputs it out of the other port on a different physical/datalink type of network. The most commonly seen media converters convert from copper Ethernet to Optical Fiber Ethernet. So this type of media converter is a small snadalone device with an RJ-45 port and a fiber optic port. It is usually intended to be attached by a short patch cable to a particular system's NIC thus converting it to the fiber optic technology. The other port can then be connected to a fiber switch port up to the maximum distance supported by the particular optical technology chosen. Alternately fiber optical cable run can reach another media converter and convert it back to copper Ethernet. This will in theory allow a system to be attached to the network even though it exceeds the maximum allowable span of the particular form of copper cabled network being used by the rest of the network. Below is a media converter that converts between 100BaseTX and 100BaseFX using SC fiber optic cable connectors:



  40. Two of these media converters could be used to patch in a single system that exceeds 100M from the central switch that services the rest of the LAN. In the following diagram all of the systems to the far right attach to the switch via 100BaseT NICs and CAT5 UTP cabling. The lone system that is farther than 100M to the left has a 100BaseT NIC but is patched across the distance by attaching media converters and allowing the signal to span the distance over optic fiber as a separate 100BaseFX segment of the network. This is a very popular method of interconnecting systems that exceed copper LAN maximum distances:



  41. Of course cabling and connectors and terminators for the buses are also part of the physical hardware of any network installation. Modern networks will either be wireless, copper cabled or fiber cabled. Wireless networks require no physical cable installation and are relatively easy to install but they still have some security issues concerning the fact that all of the network's packets can be easily captured from the air and then later decrypted. Wireless networks also offer the worst data transmission speeds of the modern networking media ranging from 802.11b = 11Mbps to 802.11a = 54Mbps ... on a very good day and in close proximity. Copper installations are more secure, but sniffing equipment can be placed near UTP cables that can record the packets which are not nearly as encrypted as the average wireless packet transmitted through the air unless all LAN traffic has been purposely configured to be encrypted by the administrator. Fiber cabled networks can easily reach Gigabit speeds, longer ranges, are impervious to EMI/RFI, and traffic cannot be sniffed by placing a device near the fiber cable. Therefore the wireless network is the easiest to install, medium in security, slowest and probably fairly limited in range if the nodes are to achieve maximum transfer rates. Fiber has the highest transfer rate between nodes, the greatest range between nodes and the highest level of physical security. Copper wired networks fall in between as far as speed, range between nodes, reliability, security and even price with Fiber being by far the most expensive network to install, older copper being the cheapest, and newer copper and wireless technologies roughly similar in price overall.

Review Questions
  1. What device is used to extend a bus network by linking two or more segments together? What OSI model layer does it occupy? It does not amplify the signal instead it does what to overcome the attenuation of the cables?






  2. Name the device also known as a multiport repeater. Explain why it is called this.




  3. List the four major types of hub and describe them:








  4. List and describe the six major components of the hub discussed in this module:












  5. List and describe the four major form factors of hubs:








  6. List and describe the four major types of bridges:








  7. Switches can be thought of as what? Explain why?




  8. Most switches function as what type of bridge at each port?


  9. List and explain the two main types of switching.




  10. What device is needed to create a VLAN? Describe a VLAN.




  11. Describe a VLAN created by a Layer 3 Switch.




  12. Describe a router.




  13. List the six main functions of modern routers.












  14. Explain the difference between static and dynamic routing.




  15. List and describe the four main routing protocols.








  16. Explain a brouter.




  17. Explain a gateway. What common network connectivity device has been expanded to gateway functionality?




  18. List and describe the four types of gateway discussed above.








  19. What is a media converter?


  20. Explain the use of 100BaseTX/100BaseFX media converters as discussed above.




  21. List six factors involved in choosing a particular type of network technology to install.




  22. If you were planning to install a new network which of the following is the least reliable (suffers the most potential signal loss or corrupted packets): wireless LAN, a copper LAN, a fiber optic LAN? And which is the most reliable?


  23. If you were planning to install a new network which of the following is the least expensive: wireless LAN, a copper LAN, a fiber optic LAN? And which is the most expensive?


  24. If you were planning to install a new network which of the following is the easiest to install at a facility: wireless LAN, a copper LAN, a fiber optic LAN? And which is the most difficult to install at a facility?


  25. If you were planning to install a new network which of the following is the most physically secure: wireless LAN, a copper LAN, a fiber optic LAN? Explain why. Explain why the others are not physically secure.




  26. If you were planning to install a new network which of the following features the greatest distance between nodes: wireless LAN, a copper LAN, a fiber optic LAN? Which features the smallest distance between nodes?




  27. If you were planning to install a new network which of the following features the highest transfer rate between nodes: wireless LAN, a copper LAN, a fiber optic LAN? Which features the lowest transfer rate between nodes?




  28. If you were planning to install a new network give three scenarios in which the best possible solution would be to implement a hybrid network involving more than one of the following types of network segments: wireless LAN, a copper LAN, a fiber optic LAN?








  29. A 10Base2 network is already up and running, what is the primary reason why the administrator would consider installing a transparent bridge?




  30. A 10BaseT network is already up and running, what is the primary reason why the administrator would consider changing the passive hub to an active one?




  31. A 10BaseT network is already up and running, what is the primary reason why the administrator would consider changing the active hub to a switch?




  32. Why would placing a bridge between a bus segment that contains all of the servers and another bus segment that contains all of the workstations be useless?




  33. You would like to reduce the total traffic on the two major segments of a network by inserting a device in between them that will filter out IP broadcasts. What type of device will be needed?


  34. Which two devices discussed above could internetwork an IBM Token Ring segment and an Ethernet segment? Which would be the more powerful (feature filled) of the two?


  35. You need to upgrade a maximum sized 10Base5 network to a 100Mbit modern networking technology. What type of media will be needed and why can't the others be used?




Copyright©2000-2004 Brian Robinson ALL RIGHTS RESERVED