In “What is a Local Area Network” we explain the basics of networking. It covers the broad technical concepts used in the rest of the presentation. “Overview of Network Technologies” takes a look at the networks available on the market today, and makes some very basic comparisons. It places the 10 Mb and 100 Mb Ethernet technologies within the big picture of LAN technologies. In “Components of an Ethernet LAN”, we identify the components which go to make up an Ethernet LAN, so that they can be explained in detail in the next two sections. “10 Mb Ethernet Technology” covers the operation of 10 Mb LANs in enough detail to discuss the design decisions, and to recognise the features in the product descriptions. “100 Mb Ethernet Technology” identifies the key differences between 10 Mb and 100 Mb technologies. “Ethernet network design” looks at the things which need to be decided when designing an Ethernet network, and gives some basic data for some of the calculations.
This section deals with the major features of Local Area Networks, and defines some of the buzz-words that are used.
A LAN is a Local Area Network . LAN usually means “in the same building”. A LAN is a collection of different technologies, not just the cable that connects the computers. The network cards in the computers, the hubs and switches in the middle, and the software that the computers run, must all work together to create a functional network.
Networks types are broadly classified by the size of area that they cover. Typically, the smaller the network, the faster it goes. Large networks which interconnect continents usually have significantly lower bandwidth than a network within an office. This is due to the cost of providing high-speed reliable services, as interference and noise become very significant over long distances, resulting in slower speeds. A Local Area Network typically connects together the computers within the same building, using copper and/or fibre cables. A Metropolitan Area Network (MAN) connects LANs together within the same district, typically using dedicated fibre optic links. Telecomm companies have been busy installing “dark fibre” in the major cities. This is fibre optic cabling without any traffic on it, which can be used as the customer wishes. This kind of service is often used for MANs. A Wide Area Network or WAN connects LANs and users together between cities and countries. Typically this uses public networks such as the telephone network, or services such as Frame Relay or ISDN. Increasingly, private operators are installing their own dedicated links; major multinational corporations have been doing this for many years, and Internet Service Providers are adding their connections to other countries.
LAN technology originated in research laboratories of the mainframe computer manufacturers, and in Universities, to connect their campus computers together. LAN technologies have moved away from being proprietary to become a standardised way to connect computers. The biggest single example of a network which connects almost every type of computer is the Internet. Without the public standardisation of LAN types, the Internet could not happen. LAN technologies are fast enough to be able to provide services that are not on the local host, such as another disk drive. Disk drives can thus be kept centrally, and protected from the typical desktop environment. This only works if the LAN is fast enough to make this process invisible.
Not all LANs are the same! We need to be able to identify their characteristics to to compare them. LAN speeds are usually quoted in bits per second : 10 Mb, 100 Mb, 155Mb, 1Gb etc. Mb, which is sometimes written as Mbps, Mb/s, or Mbit/s means “Megabits per second”. Note that this is bits , not bytes . To convert Mbit/s to Mbytes/s (Mega bytes per second), divide by eight. For example, 10 Mbit/s is approximately 1.2 MBytes/s. This is the speed at which symbols are transmitted on the wire or in the fibre, which is not the same as the end-to-end performance. The speed of a network is also known as its bandwidth . Many LANs were designed with a particular purpose in mind. They may have special features such as high reliability. Most try to be everything to everyone, but because of costs, usually provide the best value in a certain application area. For example, a technology that performs well enough to be in a critical backbone of a network is unlikely to be cheap enough to also use at the desktop. One LAN type is not “better” than another - all of them have the right area of application, and are good at different things.
Networks are based around a common principle of one cable connecting many computers together. The computers all share the same media, and only one of them can use it any any given moment. If more than one computer is sending information at the same time, their data “collides” and both transmissions are corrupted. This also means that all computers on a network can listen to all transmissions. This is known as broadcast , or multicast if the audience is agreed to be a restricted one. There are two basic ways to connect multiple computers together: on a “Bus”, where they are all connected to the same physical cable, or in a “Ring”, where a cable hops from one to the next and back to the beginning again. These topologies affect the way that the computers gain access to the network.
The way in which the computers decide who can talk at a given moment is called their Access method. The way they are connected together affects the access method. The part of the hardware that decides when the computer can use the network is called the Media Access Controller , or MAC . In the “Contention” method, the computers must be able to detect if their data is colliding with another transmission from someone else, and try again later. In the “Token Passing” method, the computers pass a token from one to the next, which gives them permission to talk on the network if they need to. There are no “collisions” on a token passing network, but it is more complex to manage, because if one of the computers fails, the ring must be reconnected around it to ensure that the token is still passed around. An IEEE standard, 802.4 (Token Bus), tried to combine Bus topology with Token Passing to get the best of both worlds. However, it remains a fringe network technology, and is not widely used in corporate networks.
Wiring computers together in a physical Bus or Ring fashion is not very practical, particularly the Ring topology. It can be difficult to connect a new computer to the network without disrupting everyone else. It can also be difficult to tell what is happening on the network, as there is no central point from which to get information. Whether the network topology is a Bus or a Ring, it is likely that it will be wired in a Star layout. The cables from each computer are brought to a central wiring point. The equipment at the centre implements the network topology. At the centre of a Star formation, there is usually a Hub or Repeater of some description - this could include switching technology and other advanced techniques. Note that it is not possible to be sure of the network type just from the fact that it is wired in a star layout. It could be either a Bus or Ring type.
In many cases, the wiring for a network costs just as much as the rest of the equipment itself. Businesses constantly undergo change, and are likely to move, change or add people and equipment. This is likely to affect the network every time it happens. If it were to involve re-wiring each time, it would become very expensive. The cost difference between installing one cable and ten cables in the same ducting is relatively small, as a lot of the cost is in the physical work. The solution is to use Structured Cabling , where the cabling for a network is installed only once, and can then be used in a flexible fashion. When a user changes their location, or when a new network host is added, the cabling is already present and can simply be plugged into the hub. UTP (Unshielded Twisted Pair) is the most common type of cable used for connections to desktops. It is described in more detail later on. If the cabling is chosen carefully, it can be used for several different types of network, so does not limit the user’s technology choices.
A node (computer) is connected to a network by adding a Network Interface Card (NIC) to the computer, and installing some Network Operating System (NOS) software. This is so that the host can both control the NIC, and run some applications that make use of the network. Each node on the network has a unique address, called a MAC Address . This address is used to identify the source and destination of a packet of data on the network. A MAC address is a set of six numbers, each of which can be from 0 to 255. It is normally written in hexadecimal, for example “00-C0-BA-00-8C-6A”. The MAC address is usually programmed into hardware, and is assumed to be completely unique. Each manufacturer of network equipment has a special code (called an “Organisationally Unique Identifier”, or OUI) which is used in their MAC addresses to ensure that they are never the same as anyone else's. A typical network application is to make a disk drive with programmes and data on it appear on the local machine. Physically, the disk exists on a remote computer (a “server”), but the network and software make it appear to be a part of the local computer, typically called a “client” in this context. Some computers already have network interfaces installed in them. If not, most have an expansion bus of some description which can take a plug-in card. Many operating systems already have Network capabilities built into them. For those that do not, networking software can be added.
If we compare the hardware of a computer network with the road system, then Protocols are the “Highway Code”, or the rules describing how we should drive on the roads. Having common rules means we can drive from one town to another using the same techniques. If neighbouring towns drove on opposite sides of the road, and had different meanings for the traffic lights, travel would be hazardous. With computers, using agreed methods of interchanging data on a network allows computers from different manufacturers to talk with each other, without any special provisions having to be made. Some protocols are in the public domain, and are not owned by any one company. Examples are TCP/IP (the protocol used on the Internet), and OSI (the protocol standardised by ISO). Other protocols are owned by companies, but are still publicly supported by other vendors. Examples include IPX, the protocol used on Novell Netware networks, and Appletalk, Apple Computer’s network protocol. The network hardware is only concerned with delivering packets of data from one point to another. It is not concerned with what is in the packets. The protocol determines what is carried in the packets, and how it is handled at the other end. Some protocols also ensure that the data is correctly delivered, by checking that the right number of packets have been received, and that they are put back into the right order again before use by the application software.
If the users on one network need to access stations on another network, the networks must be connected in some way. There are three basic methods of joining network segments together; repeaters, bridges and routers . The ISO model for networking defines seven layers of network implementation. They are numbered from the physical parts of a network through to the final applications that the users see. For the purposes of this discussion only the first three layers are mentioned: Layer 1: The Physical Layer, which is the cabling and transceivers Layer 2: The Data Link Layer, which describes the frames of data moved around between network nodes, using physical network addresses Layer 3: The Network Layer, which describes the ways in which network protocols communicate with each other, with the data that is contained in the “payload” of the frames Repeaters connect network segments at layer 1. Bridges connect networks together at layer 2, and Routers interconnect networks at Layer 3. These functions are described in more detail in the following slides.
Connecting two network segments with repeaters joins them together at every level. All traffic is visible to all nodes on both segments. Repeaters do not interpret the contents of the network frames at all. Only network segments of the same type can be joined together with Repeaters, and only if the resulting network obeys the architectural rules of the network topology. Repeaters add a small amount of delay to the network, which can be important in high-speed networks.
Bridging two networks, by connecting a Bridge between them, makes one large network. All the hosts are now visible to each other, and all types of traffic flow freely across the bridge. When one station sends a broadcast packet, it is also broadcast in the other network. Bridges only need to look at the first part of datagrams that they receive, and will forward them to the other network if the destination MAC address belongs to the other network segment. Learning Bridges learn the MAC addresses of the nodes on either side of their connections by reading the source MAC addresses of frames arriving on their inputs. Most network packet formats contain the source address of the sending station in them. (Ethernet does, as we will see later). Bridges can only be used to connect networks of the same type together. They can be used over a WAN to join a remote user to a corporate network, for example with a dial-up connection over ISDN. Because bridges have to look at the destination MAC address within a frame before knowing whether to forward or ignore it, they add some latency to the network and are slower than Repeaters.
Routing between networks is a little more sophisticated. A Router , instead of simply copying everything from one network to the other, will only send datagrams which are specifically addressed (at the protocol level) to hosts on the other network. It does this by receiving the datagrams, interpreting them, and converting them if necessary before sending them on. Routing between networks keeps the networks separate, while allowing controlled communications. It is used where security is an issue, and where the cost of sending data between networks may be significant (e.g. over long distances). Routing does not work at the basic packet level, but understands the protocol inside the packet, as some “routable protocols” contain information about the intended route inside the packet. Routers also connect different network types together, because they transfer information at the protocol level, not the frame level. For example, an Ethernet router can take Internet Protocol frames and send them to a router on a Token Ring network, which will convert the frames into Token Ring packets. Routers also have another advantage over bridges; it is possible to have multiple paths between networks, so that if one path fails, another one can be used. Bridges cannot do this, but Routers are able to provide this level of redundancy. Because routers have to receive and possibly translate the complete contents of a frame, they add latency to a network, and are slower than bridges.
In this section, we have looked at Why networks are used and what they do Network topologies (bus and ring) Access methods (token passing and contention) Star wiring topology and structured cabling Bridging and routing The use of switching to improve overall throughput Segmenting networks for management, security and performance The concept of Virtual Networks In the following sections we will refer back to these concepts.
In this section we will briefly review the major network technologies available on the market, so that Ethernet can be positioned against the other solutions. There are at least seven major “open” network technologies available, with several other fringe technologies that have not made significant impact on the market. By “open”, we mean that equipment is available from more than one supplier, and equipment from different vendors will interoperate. Proprietary technologies are not covered in this analysis, nor are network protocols (the language that computers use to transfer data).
Ethernet, Fast Ethernet and Gigabit Ethernet are basically the same technology scaled up in speed by a factor of 10 each time. Some small adjustments have to be made to accommodate changes in timing. Equipment which supports more than one speed (N-Way) is available, so to some extent a system can be upgraded in steps to take advantage of higher speeds. However the main advantage of considering 100 Mbit/s Ethernet as an upgrade path from 10 Mbit/s is familiarity with the network architecture and operation. The cost per Mbit/s of the Ethernet family is also the best value of all networks, due to the well established nature of the base technology. This has minimised the additional investment needed for new developments, when compared (for example) to a completely new approach such as ATM. Ethernet technologies are likely to continue to dominate the corporate data LAN, although other technologies will still prevail in WANs and other areas, such as demanding industrial or military environments. New communications protocols coupled with cheap bandwidth will make multimedia (voice and video) cost-effective over the corporate Ethernet LAN.
Gigabit Ethernet, operating at 1000 Mbit/s, is set to challenge 622 Mbit/s ATM as a high-speed backbone technology. This means that large hubs and routers can be linked with very high speed connections, where the aggregated traffic between large networks needs to be carried over short distances. Gigabit Ethernet to the desktop is unlikely to be a significant market until computing power and the capacity of expansion buses makes it practical. Today, a good PC or an entry-level workstation may be capable of handling 3 to 6 Mbytes/s at a network port. Gigabit Ethernet carries over 100 Mbytes/s, so there are few computers that can exploit this as a network interface (yet). When coupled with the cost of a Gigabyte Ethernet port on a switch, the relative benefit when compared to Fast Ethernet is negligible for all except very high performance servers. The technical challenge of carrying 1000 Mbit/s over copper cable is still being researched in development laboratories. The problem is that as signal frequencies increase, the amount of “cross-talk” between wires in the same cable or trunking also increases. At 1000 Mbytes/s, the frequencies are well above usual radio or television, and a piece of wire only a few millimetres long makes an efficient aerial!
Network Technologies can be grouped by speed: under 100 Mbit/s, 100 Mbit/s and above. Until recently, many users accepted that 100 Mbit/s was unnecessary for connections to desktop machines, but as processor speeds and the demands of applications continue to increase, this limit is less justifiable. The technologies can also be grouped by Access method: token passing or contention. In general, token-passing technologies have somewhat better performance under heavy loads because there are no collisions, but are costlier to implement and manage. Contention-based networks have better performance under conditions of lighter loading, because there is little delay in accessing the network. To date, the “killer application” which mandates the use of deterministic networking on a large scale has not emerged, and it seems that the technical challenges of delivering real-time multimedia will largely be met by simply increasing bandwidth, rather than deploying different network technologies. While other network technologies have their specialities, Ethernet is the one technology that covers the different speed requirements of all parts of a LAN, from desktop to backbone, and offers the lowest cost/performance ratio. This is the reason that Ethernet has dominated in the LAN and is likely to continue do to so.
In this section we take a more detailed look at 10 Mbit/s Ethernet, in preparation for preparing to design Ethernet networks. The functions of the components typically found in 10 Mbit/s Ethernet are explained in general terms.
Ethernet is named after the “Luminiferous Ether”, an invisible substance through which electromagnetic radiation was once thought to propagate. Today’s “Ethernet” is DIX Ethernet II. The IEEE adopted what was a de-facto standard as the basis for a public standard in 1985, and termed it IEEE 802.3, but not before they changed the packet format. While IEEE 802.3 and Ethernet II are compatible at the physical level, they are incompatible in terms of the packet format. Only ISO protocols use strict 802.3 formats; Ethernet products use the Ethernet packet format. Ethernet uses Baseband transmission, which means that it uses all of the cable to carry a single channel of information (the network traffic). By comparison, Broadband operation carries multiple channels of information in the same media - radio transmissions at different frequencies are an example of Broadband operation. The various types of Ethernet, and some other network technologies too, are described in a “ Speed -BASE- Media ” format. Speed represents the network speed in Mbit/s, and Media indicates the type of cabling used, or sometimes indicates the maximum transmission distance. For example, 10BASE-T means “10 Mbit/s, Baseband transmission, T wisted pair cable”. This method of describing the capabilities of the network is widely used and is referred to in the following slides. The term “10BASE-X” refers to the whole family of 10 Mbit/s technologies.
The basis of Ethernet is the Contention-based mechanism for sharing the cable between multiple users. This requires that nodes connected to the network are able to listen to what is on the network while they are also transmitting. It also requires that if two nodes transmit at the same time, then the resulting pattern on the network is detected as being corrupt. When two (polite) people are talking, and they both start to say something at the same time, one will usually stop and allow the other to continue. This is the issue that Ethernet solves, which allows many machines to share the same cable. By comparison, some other network technologies use CSMA/CA , which is Collision Avoidance rather than Collision Detection . In CSMA/CA, the transmitting station sends a warning before it transmits, and other stations are expected to observe this signal. Contention-based mechanisms such as CSMA/CD are also called probabilistic networks, because the average delay that a station will experience in accessing the network is a function of probability, and is not guaranteed. Contention works well with traffic levels up to around 30% of network speed, after which collisions have a significant effect on the available bandwidth. Above about 70% of network bandwidth, almost no useful data is delivered.
All the steps in CSMA/CD are controlled by the NIC or Adapter card hardware - the software is not really aware of this issue, although it may be able to collect some statistics about network collisions. When a station has data to send , it first listens to the network to see if the network is in use. This is the Carrier Sense part of CSMA/CD. If there is no traffic on the network, the station can start sending data. The NIC card will prepare the data for sending as ‘packets’ and start to transmit them on the network. It is free to do this without any other consultation with the other nodes on the network - this is the Multiple Access part of CSMA/CD. If another station had decided to do this at exactly the same time, the data from the two sources will ‘collide’ on the network cable. Ethernet NIC cards can sense when this happens by the voltage increase in the cable caused by two transmissions (just as you can tell when the radio volume goes up). This is Collision Detection of the CSMA/CD system After a collision occurs, all station s on the network will cease transmitting and will wait a random period of time before attempting to transmit again. Because the time period is random, the probability that the same two stations will restart transmissions at the same moment is quite small. The first station to reach zero on its random timer will begin transmission again, after first checking to see if the network is quiet. No single station can monopolise the network, as there are maximum packet sizes that can be sent, and they can only be sent one at a time.
The discussion of CSMA/CD in the preceding slides only applies to Half-Duplex circuits. A half-duplex circuit is one where there is only one circuit for carrying data. From the point of view of any individual station, this single circuit is used for both transmit and receive. When the station is transmitting, it controls the circuit. When it is listening, it must temporarily “disconnect” its transmitter, to allow another station to control the circuit. Full Duplex describes the ability of a station to both send and receive at the same time , so Half Duplex means that the station can either be sending or receiving, but not both at the same time. Half-Duplex applies to connections where the media is shared, such as the “Thin” or “Thick” coaxial cables, or connections made through a multi-port repeater (wiring hub). All stations connect to the network in “parallel” and share the same media. Repeaters cannot support full-duplex operation, because if they detect transmit signals from two sources at the same time, they will interpret this as collision and will “jam” the network to ensure that all nodes observe the collision.
Full-Duplex applies to connections where there are only two stations at either end of a link, which contains separate receive and transmit circuits. When a computer is connected to a hub with a twisted pair wire, for example, there is little point in having a half-duplex cable, because it does not really simplify the cabling. The cable can quite easily have two pairs in it, one for send and one for transmit in either direction. This means that the station can be both sending and receiving at the same time. Note that in the case of only two stations on a shared media connection (e.g. 10BASE2), this does not mean that they operate in Full Duplex. There is still only one circuit in a 10BASE2 connection. Also note that Full-Duplex can only operate from one station to another, over the complete link including any hubs. If there are any shared-media segments in the path between the stations, Full Duplex cannot be used.
Data of all types is sent over networks in small pieces. There are two main reasons for this. First, this shares access to the network more fairly. When a packet of data has been sent from one computer, then another one gets a chance to use the network in the gap before the next packet is sent. If a network node were allowed to send all its data at once (e.g. a 100 MByte file), then none of the other nodes would be able to access the network while it was doing this. Second, there is an amount of data which is “safe” to send in one packet. To ensure that the data arrives in the same condition in which it was sent, a checksum is added to each packet. This number is the equivalent of all the rest of the packet data added together, and is recalculated at the receiving end. If the checksum does not match on a received packet, then there must be some kind of corruption or loss, and the packet is not safe to use. Corruption can be caused by electrical interference or a weak signal. The bigger the packet, the greater the chance of corruption occurring somewhere within it. For example, if our 100 MByte file was sent as one packet, the probability of it being received in perfect order on a typical network is virtually zero. However, the chances of smaller packets (up to 1500 bytes in the case of Ethernet) being received correctly are close enough to 100% on a correctly installed network to make this a practical upper limit for the packet size.
It is important to understand the differences between the forwarding algorithms that Ethernet switches use. Prior to this, we need to review the structure of an Ethernet frame. The first eight bytes of the frame are the preamble ; this identifies the frame as valid Ethernet. After the preamble comes the destination address , the six-byte MAC address of the station to which the frame is being sent. The special address (in hexadecimal) of FF-FF-FF-FF-FF-FF is the broadcast address, which means that all stations on the network are to receive this frame. Next, the six-byte source address identifies where the frame is coming from. The source address is the key to a switch’s ability to build up a forwarding table. At the end of the frame, there is a checksum or cyclic redundancy check (CRC). The transmitting station performs some arithmetic on the entire frame just prior to sending it and puts the result in the CRC field. The receiving station performs the same arithmetic on the data that it has just received; if the result is different, there must have been an error in the transmission. Frames can be between 64 and 1,516 bytes long. This means that the minimum amount of data is 46 bytes, as the fixed part of the frame is 18 bytes.
In this section, we will look at the roles played by the various components of an Ethernet network. In the diagram above, the NIC is shown as connected to the Hub either directly if it has an integral transceiver (XCVR), or via an external transceiver if not.
The Network Interface Card, often called the “Adapter card”, is the physical link between a computer and the network. Usually, the NIC handles the low-level time-critical parts of the network system, such as controlling access to the network, and actually sending and receiving the frames. It is unusual for NICs to understand the data that they transfer, although there have been “protocol processing” adapters on the market. The NIC usually contains a MAC address built into the hardware. As previously mentioned, this MAC address is guaranteed unique, and is programmed by the manufacturer. It is unusual to need to change the MAC address, but can often be done through a software utility. Some protocols may need MAC addresses to remain consistent, so if a card is changed, the MAC address may have to be restored. However the majority of protocols in common use will adapt automatically to a new MAC address. With high-speed networks and high capacity servers, the performance of the network card itself becomes a critical factor in network throughput. The I/O bus into which the NIC is plugged is also a factor. This subject is examined in more detail in section 5.
Transceivers are used to convert electrical signals from and to the Ethernet. Originally all devices had an AUI port (Attachment Unit Interface). The AUI output was in the right packet format but still needed conversion to Ethernet electrical signalling. Today most NICs and other equipment has this functionality built in so that the traditional transceiver requirement has to some extent disappeared. Most NICs offer direct 10BASE2 or 10BASET connections. However, in 100 Mbit/s Ethernet, the interface between the transceiver and the “DTE” (Data Terminal Equipment) is also standardised, and is known as “MII” (Media Independent Interface). This is explained further in the section on 100 Mbit/s Ethernet.
Routers and switches often still use AUI ports, so transceivers are necessary to make a connection to 10BASE2 or 10BASE5 cabling. Older types of NIC may also have an AIU interface.
Ethernet cable comes in many types. The original cable was thick coaxial cable like a very large TV aerial cable. This was expensive and difficult to wire up, and was known as “Thick Ethernet”.The connector for this were called N-Series (which was a screw in arrangement) and Vampire tap ( which used a sharp needle to connect to the cable). Thin Ethernet (called 10BASE2) uses cable just like a TV aerial and connects using the BNC connector. It was much easier to wire and is still a good option for very small networks. 10BASE-T cabling is the most common today. This uses cable very similar to telephone cable and is cheap and easy to wire. All UTP systems require hubs (we will look at hubs a little later). Finally, for people who require security or long distance or have noisy electrical backgrounds (such as manufacturing plants or airports), Fibre Optic cable is now cheap and widely available. In fact Fibre represents a very low cost solution today has the benefit of being able to carry very high bandwidth so that it is ideal for future proofing a network cable plant. Electrical noise is like radio interference where the signals in the cable are disrupted by powerful electric equipment, for example when you get radio interference from power tools.
The 10 denotes that the Ethernet is a 10 Mbps system (fast Ethernet is prefixed by 100 indicating a 100 Mbps system). BASE indicates that the network is Baseband, i.e. there is only one signal on the network. In the early days of LANs there were networks that operated BROADBAND systems where there were multiple signals on the cable and these were tuned in or out by the transceiver. The final character denotes the maximum distance for the cabling scheme in hundred of metres. 5 - 500 metres for thick coax 2 - 200 metres for thin coax (in fact the maximum is 185 metres) T - Indicates twisted pair cable which is 100 metres maximum distance FL - Fibre Link You should be aware that there are different grades of twisted pair cable for Ethernet. The category number of a cable is related to the number of twists per foot in each pair. Higher numbers of twists will support higher frequencies in the cable without excessive cross-talk (interference with other cables). Category 3 is voice-grade cable, commonly used for telephone circuits. Category 5 is data-grade cable, required for 100 Mbit/s operation. Categories 3, 4 and 5 will all support 10 Mbit/s operation.
A Repeater is used to connect together segments of Ethernet cabling, to extend the physical reach of the network. It amplifies and re-shapes the Ethernet signals so that they can be clearly received. Repeaters understand the contention-based access mechanism, and re-implement it. Everything connected to a repeater, whether a shared cable or a point-to-point cable, is included in the shared media. The repeaters reconstructs the electrical and timing definitions of the frames, and detects collisions. When a collision is detected, a “jam” signal is output to ensure that everyone on the network also sees the collision.
It would be useful if repeaters allowed us to build infinite networks. But because of the signal delays introduced by long cables, there are some rules limiting the design of Ethernet networks, known as the “5-4-3” rules. These say that between any two stations in an Ethernet LAN, there should be a maximum of five segments four repeaters three segments with multiple stations connected to them. (The “segments” described in these rules refer to lengths of cable within the same network which are connected by repeaters. They do not refer to “segmented networks” which are connected by bridges or routers). A network which broke these rules would be likely to exceed the maximum “propagation delay”. This means that a if two stations at the opposite ends of an oversized LAN decided to transmit at the same time, then they may not detect each other’s signals early enough to clearly identify a collision, because it would take so long for the signals to reach the other end. This is described in more detail in the following slides.
The Round Trip Delay Time is the time that it takes for a signal to go from one end of the network to the other, and back again. The worst case is when a station at one end of the network sends the shortest allowable frame (64 bytes x 8 bits per byte = 512 bits). To ensure that the station at the very far end of the network sees this frame by the time half of it has been transmitted, then the outward trip must be done in 256 bit times. The round-trip (there and back) must therefore be completed in 512 bit times. Some example calculations of round trip delay times, and the network components which influence them, are given in later slides. This applies to 10BASEX, 100BASEX and also to Gigabit Ethernet. In the case of Gigabit Ethernet, the transmission speed is so great that some additional limitations have had to be added to the minimum size of packets. The alternative would be that Gigabit Ethernet networks would have to be so small that they would be impractical.
There are three basic types of hub. Hubs are used at the centre of Star wiring configurations in Ethernet networks. Passive Hubs simply act as a wiring box for the network. They link the cables from the attached stations together to make up the network. Because the hub does not amplify the signals in any way, it must be taken into account when considering the 5-4-3 rule. Active Hubs regenerate the signals so that longer cables can be used. They are also known as Multi-port Repeaters . Intelligent Hubs are also Active Hubs, with additional functionality added. Typically, an intelligent hub includes management functions which allow the network administrator to configure and monitor the system. Intelligent hubs may also include switching, which sends specific data direct to the recipient instead of broadcasting it to the whole network.
The advantages of using hubs and Star wiring are many. The control and monitoring functions of the network can be brought together in one place, and direct connections between hubs can be implemented easily. Many hubs can support multiple media types and speeds, so that cable types can be optimised for the physical area they serve, and mixed in the hub. Intelligent Hubs which support Virtual LANs allow the network manager to configure and control network security, so that sensitive areas can be protected from general access.
The descriptions here are not complete but serve as a comparison between techniques used to interconnect Ethernet networks. If a repeater is used to connect segments, the two segments become a single collision domain, and the 5-4-3 rule applies to the network design. If a Bridge is used to connect networks, the segments remain as individual collision domains. Only frames which are addressed to the other segment are copied across. Switches act as multi-port network bridges. There are several techniques, such as store-and-forward, cut-through and fragment-free switching. Switches can also change network speed between segments, assuming the frame type is the same.
In this section, we compare 100BASE-X Ethernet technology with 10 Mbit.sec Ethernet. 100BASE-X is also known as “Fast Ethernet”, and is the net generation of Ethernet products. Not all the previously described principles are covered again in this section, only the areas where 100BASE-X differs from 10BASE-X.
100BASE-X uses the same CSMA/CD contention-based technology as 10BASE-X. However, there are no multi-drop cable options in 100BASE-X, so everything is wired through hubs or repeaters. All links in 100BASE-X are point-to-point only. Collisions still occur, but only inside multi-port repeaters (wiring hubs), so there are still some rules about network design that limit the size of networks. 100BASE-X uses the same frame format, so that all software remains unchanged. It is easy to update from 10BASE-X to 100BASE-X. Parts of a network can be upgraded to eliminate bottlenecks, without having to upgrade the whole system.
Since 100BASE-X Ethernet is an upgrade path from 10BASE-X, there are many cases where both technologies are used in the same network. Some equipment may support both 100 Mbit/s and 10 Mbit/s operation at the same port. In order to ensure that the network is plug-and-play, a scheme called “N-Way autonegotiation” exists to resolve any differences, and establish a connection. Auto-negotiation resolves two parameters; speed and duplex. If 100 Mbit/s is available at both ends of a link, then it will be used. If full-duplex operation is available at both ends, then this will also be enabled, otherwise half-duplex operation is configured. Both devices must support N-Way autonegotiation for it to detect duplex.
MII, the Media Independent Interface, is an updated version of the10BASE-X AUI interface. It defines a connector type and specific functionality. When a DTE (Network Interface Card or hub port) is connected to a 100BASE-X transceiver through a MII connection, this gives the DTE the ability to control the functions of the transceiver, and also receive information from the transceiver about its abilities. This is used in N-Way autonegotiation. The same adapter card (for example) can be used with a variety of transceivers, and can make the best use of each.
All media options are for point-to-point links in 100BASE-X. 100BASE-TX is the equivalent of 10BASE-T. (UTP = Unshielded Twisted Pair). 100BASE-FX is the equivalent of 10BASE-FL. 100BASE-T4 has no equivalent in 10BASE-X, and is described in more detail in the following slides. There is no 100 Mbit/s equivalent of the multi-drop 10BASE2 and 10BASE5 cables, so copper cable distances are more limited in 100BASE-X.
100BASE-TX is a point-to-point link for 100 Mbit/s Ethernet. As it uses two pairs, full or half duplex operation is possible.
100BASE-FX uses fibre optics to make a point-to-point link. It uses the same fibre optic cabling as FDDI and ATM, so a 62.5/125 fibre pair installed as part of a structured cabling scheme can be used for different applications as necessary. Since there are separate receive and transmit paths, full duplex operation is possible, but 10 Mbit/s is not supported, so auto-negotiation is not relevant. Fibre optics: Multimode fibre optics uses lower power LED transmitters, and is affordable. Single-mode fibre optics uses laser transmitters for very long cable distances (up to 20 km), and can be very expensive by comparison. The term “62.5/125” refers to the dimensions of the fibre optic cable, in microns (1 millionth of a metre). A fibre optic cable is made up of an inner glass core of one refractive index, and an outer layer of a different refractive index. The change in refractive index forms a reflective surface, off which the light bounces as it is guided down the fibre. The inner core is 62.5 microns in diameter, and the outer layer is 125 microns. It is important to match the fibre optic cable to the transceivers and plug types, as they are an exact fit to ensure minimum losses. Single-mode fibre has a different specification.
Since it is not possible to extend 100m cable lengths by adding repeaters, the only real use of multi-port repeaters is to add more users to a network. (The distance limitations are explained in a subsequent slide). 100BASE-X repeaters perform the same functions as 10BASE-X repeaters in terms of cleaning up the contention-based access mechanism. However, since there are no shared cables defined in 100BASE-X, multi-port hubs are also the place where the media is shared, and the collisions take place.
The delay introduced by a 100BASE-X repeater is important, since it affects the design of the network. Delays are quoted in “bit times”, since the basic rule of a maximum round-trip delay of 512 bit times still applies to 100 Mbit/s Ethernet.
If Class I repeaters are used, there can be no more than one repeater between any two nodes. This means that a Class I repeater cannot really be used to extend the physical reach of a network - it can only be used as the hub of a network. Class II repeaters can be connected, because they are faster. However, they cannot have more than 5 metres of cable between them, so like the Class I repeater, they cannot be used to effectively extend the reach of a network. They would typically be used in a rack at the network hub, with patch cables joining them up.
The delay in bit-times for the various components of 100 Mbit/s Ethernet are available, so it is straightforward to calculate the round-trip delay time of a particular collision domain . This is necessary to ensure that the network will correctly detect collisions. Since 100 Mbit/s Ethernet runs ten times faster than 10 Mbit/s, and the permissible round trip delay times are quoted in bit times, it follows that allowable delays in 100 Mbit/s Ethernet will be smaller in terms of absolute time. This means that 100BASE-X networks tend to be physically smaller, unless fibre optics are used.
The delay times vary depending on the media used. The key to the above table is: “TX” is “100BASE-TX”, the UTP Category 5 interface “T4” is “100BASE-T4”, the UTP Category 3 interface “FX” is “100BASE-FX”, the fibre optic interface “Shielded Twisted Pair” is 100-ohm foil wrapped cable, not the IBM Type 1 cable (which is also known as STP). “DTE” is “Data Terminal Equipment”, the computer or end-station connected to the network A margin of 4 bit times should be added to any calculation to be on the safe side.
In this example, the round trip delay time is greater than 512 bit times, so the network will not function reliably. To correct this, it would be necessary to reduce the cable lengths or perhaps re-design to eliminate one of the repeaters.
The example above is close to the 512 bit time limit, but is functional. If a Class I repeater had been used, it would be outside the design limits.
In order for an intelligent switching hub to support 100 Mbit/s Ethernet, it must have a high backplane capacity. If 10 Mbit/s ports are also supported, the hub must be able to do store-and-forward switching. A typical application for a high-capacity switch would be to provide 100 Mbit/s links between the backbones of other networks.
100 Mbit Ethernet is a logical step forward for 10 Mbit/s Ethernet installations. It can be added in increments, by using the N-Way Autonegotiation feature to mix 10 and 100 Mbit/s equipment. 100 Mbit/s links can use existing Category 3, 4 or 5 wiring, but does not use any shared cables. Since bit times are much shorter in 100 Mbit/s Ethernet, it is essential to perform the round trip delay time calculations, so ensure that the network will operate reliably.
Because of the relatively short distances which 100Mbps Ethernet can run, the need for switches to extend the network is much greater. Switches perform a similar function to traditional repeaters in 10Mbps Ethernet for Fast Ethernet networks. Because of this Fast Ethernet is sometimes described as Switch-Centric (ie switches are a central point of any design). The falling costs of 100Mbps switches has made this viable for even small companies. In the illustration the network is within limits for 7 switches/bridges between nodes and the link distances between hubs and switches could be 100Metres of copper, switch to switch could either be 100meters of copper or 412meters of fibre allowing for a large network diameter.
When all the computers on a network share the same media, the maximum total amount of traffic that can be flowing at any one time within the network is limited to the capacity of the media. In the diagram above, we have eight computers. Let us assume that the network they are on is a 10 Mbit/s network. Each of them has a connection to the network. Most computers on the market today are capable of running their network interfaces at 10 Mbit/s, or 1.2 Mbytes/s. (This has not always been true, and is still not true today for 100 Mbit/s technologies, which require a lot more processing power to send and receive network data at 12 Mbytes/s.) This means that at any given moment, the maximum amount of traffic that can flow in the above network is 10 Mbit/s. Further, only two computers can be exchanging data with each other (ignoring broadcast traffic). However, as there are eight computers, it is possible that up to four simultaneous conversations between pairs of computers could be required at any given moment. The maximum amount of traffic that the computers could generate between them is 4 x 10 Mbit/s, or 40 Mbit/se. We have the situation where the network is capable of carrying 10 Mbit/s, and the computers on it are capable of generating 40 Mbit/s. The network becomes the bottleneck, limiting the work rate of the computers instead of accelerating it. Switching technology can be used to overcome this.
Instead of simply connecting all the computers to the same media, we can connect them to a Switching Hub instead. A Switching Hub re-directs user data to the intended recipient alone, instead of the data being sent to everyone on the network. (This excludes broadcast traffic which is intended to be sent to everyone). It does this by examining the destination MAC Address of the data, and knowing where that host is connected to the hub. A “private” conversation is then allowed between two computers. A Switching hub can have a “backplane capacity” which is far greater than the bandwidth of any of the individual connections. It can therefore support many simultaneous conversations between pairs of attached computers. The total amount of traffic on the network at any given moment is not limited by the capacity of the cable (e.g. 10 Mbit/s in the earlier example). It is instead limited by the hosts themselves, or the backplane capacity of the switch. Switches do introduce a short delay, because they need to read the destination address of a packet before deciding where to send it. This is covered in more detail in later slides. Computers connected to a Switching Hub are more likely to get the most out of their network connection, as they will spend less time waiting for access to the network.
As more and more users are connected to a network, the performance goes down in a non-linear fashion; typically, doubling the number of users would result in less than half the previous overall throughput. This is due in part to the effects of broadcast traffic on the network, which is an “overhead” needed to keep the information about the hosts up to date. On a large network, the majority of exchanges will be between well-defined areas. Traffic from any one host to all the others is not likely to be evenly distributed; most users will communicate with only a few other computers. There will be other computers with which they will never need to communicate, or at least only very rarely. This leads us to conclude that it is inefficient to provide the same bandwidth and performance for both frequently and infrequently used connections. It would be better to ensure that the frequent connections have the best performance, and tolerate a lower level of performance for the infrequently used ones. Segmenting a network by dividing it up into a number of smaller networks is the most cost-effective way to achieve this goal. Network users can be grouped into functional segments, not necessarily physical ones; Two groups of users may be collaborating on a project but may be in separate buildings.
Although the diagram shows a star-wired network with a hub at the centre, the same principles are true for any network type. The more stations on the network, the more broadcast traffic there is. Broadcast traffic is not usually user-to-user data; it is used by the network itself to discover and update the locations of computers on the network. Although it is usually invisible to the users, it still takes up bandwidth. When a hub is used at the centre of a network, it has to forward broadcasts from every station to every other station on the network. This means that the amount of broadcast traffic goes up with the square of the number of stations. If the number of stations in a network is doubled, the amount of broadcast traffic that the hub has to forward goes up by a factor of almost four. With large networks, this load can become significant. Managing large networks is also more complex. The conclusion is that smaller networks give better performance to their users, even if switching technology is used.
Most large networks can be segmented into smaller sections, based on the user groups served. The objective of dividing a network is to minimise the amount of inter-network traffic that flows between networks. For example, if the Accounting Department and Shipping Department shared the same network, there would be a lot of traffic being carried which was of little use to both departments. Segmenting their networks into two smaller ones means that each is more efficient, and the link between them only carries data from one department to the other, as broadcasts need not be sent between networks. The amount of broadcast traffic within each of the network segments is also greatly reduced, as the number of computers in each segment is halved. The result is two smaller networks which are more manageable and give better performance to their users.
We looked at Switching techniques in general in an earlier section. The next few slides add some more detail to this. To summarise, Switching hubs direct traffic between individual stations, rather than copying everything to every port. This allows simultaneous conversations to take place, which increases the overall utilisation of the network. Switching is done on the basis of MAC address, which is the unique hardware address for each station. Switches learn the MAC address of the attached stations by reading it from frames received from that port (the Ethernet frame contains both the source and destination MAC addresses). It makes a note of the MAC address of the attached stations in a table, and when a frame is received for that specific MAC address, it is sent to the right physical port. Since there may be a multi-drop cable segment on a port, there may be many MAC addresses assigned to any port. They can also change, as stations are added and removed, so the switch must keep its lists up to date. Frames which have the Broadcast bit set, or which use the reserved Broadcast address of FF-FF-FF-FF-FF-FF, are sent to all ports.
There are several techniques used for Switching Ethernet frames, described in the following slides. The first, Store and Forward , means that the whole frame is stored in memory in the Switch and then sent out to the destination port. Store and Forward would be used where the outgoing link runs at a different speed to the incoming link. For example, if a frame is received on a 100BASEX link, it cannot be simultaneously copied to a 10BASEX link, because of the speed mismatch. It must be stored first and then sent out at 10 Mbit/s. This adds to network delay. (Note, however, that this is not the same as propagation delay, because the input and output links are two separate network segments. Delay is seen in the overall delivery of the packet, but this does not involve collision detection on the two segments). The second technique is to use Cut-Through switching. Here, frame forwarding starts as soon as the destination address has been determined. This saves on the amount of memory used needed for temporarily storing data, as only the first few bytes need to be stored. It also minimises the delay introduced by the switch. However, cut-through switching can only be used between network segments running at the same speed - it cannot be used between different speed segments.
An additional switching technique is Fragment Free cut-through . Here, frame forwarding starts only after the first 64 bytes of the incoming frame have been received. This helps to ensure that only valid and complete frames are forwarded. In Ethernet, frames can be aborted due to collisions at any point up to the 512-bit time. On any network using contention, there will be some “fragments” of frames which were never finished. Forwarding these onto another segment has no purpose, as they do not carry any usable data unless they are complete with a CRC. If the switching hub waits until the 512-bit time has elapsed (64 bytes), then it knows that this is a valid frame and is worth sending on to the recipient. Fragment-free cut-through adds to the delay time introduced by the hub, but if the attached segments have multiple nodes, then there is a benefit in not repeating the collisions from one segment onto another.
In the next group of slides, we look at some of the ways that switches can be used to construct networks. One of the considerations in choosing a switch for a particular application is the MAC address capacity. The switch must be able to store the MAC addresses of all the nodes on each port, so the total number of nodes on the network supported by the switch cannot exceed the MAC address capacity of the switch itself. In the first example, Segment Switching, a switch is used to interconnect a series of smaller network segments. These network segments are themselves based around simpler hubs. The switch acts as a kind of router between the segments, only forwarding frames which are destined for MAC addresses on other segments, and not repeating other traffic. (However, it differs from a proper router in that it operates at the MAC address level, and is independent of the protocol used). This allows each of the smaller network segments to use a separate Ethernet network, with connections to other network segments being made on demand through the switch. It is more efficient than simply connecting all the nodes together in a large network, as network utilisation (and efficiency) is optimised.
In the second example of switch usage, we have a series of workstations directly connected to a central switch. Because there is only one node per port (instead of a complete multi-node network segment), the MAC address capacity of the switch does not need to be large. Each of the workstations can hold conversations with any of the others, making use of the full bandwidth of their connection. This layout would be suitable for a small group of users with high bandwidth demands.
The third example is a scaled up version of the first, where a switch is used to interconnect what are already substantial networks. The main features of a switch needed to support this type of application would be a very high “backplane bandwidth”, and a very large MAC address capacity. Switches can be constructed in a modular fashion, so that they can scale with the application.
The last application example for switching is to use a switch as a “bandwidth converter”. A switch can have multiple 10 Mbit/s ports which support either network segments or individual power users, and 100 Mbit/s port or ports for connection to another part of the network. The switch can aggregate the 10 Mbit/s traffic for transmission at 100 Mbit/s between network backbones. This eliminates the traditional inter-network bottleneck, allowing traffic to flow at higher speeds in critical parts of the system to compensate. To do this, the switch must support store-and-forward features (as previously described).
A switch cannot really be described as a bridge, a router or a repeater in traditional networking terms, as it performs some aspects of each. In terms of the 5-4-3 rule, a switch can be regarded as a bridge, because it receives and re-transmits frames. Repeater and segment counts can be restarted on the other side of a switch. Because bridges (as opposed to repeaters) add delay into the network, a practical limit of 7 bridges is proposed in any network. The 7 bridge total however is derived from Token Ring networks which have a finite limit of 7 bridges between nodes. The IEEE 802.3 commitee decided to limit Ethernet to the same bridge rules for simplicity. Although this number is purely arbitary for Etherne the number of switches between any client and the servers should be kept low, to avoid unnecessary delay. Significant latency in the network will restrict transfer rates, as some protocols (e.g. TCP/IP) will only allow a limited amount of data to be transmitted before an acknowledgement is received. If there is a delay in arrival of the acknowledgements, this restricts the overall transfer rate.
Using switching techniques and segmenting networks can lead to big improvements over large shared-media networks. However, it is unlikely that any corporation could decide on their optimum network layout once, and leave it unchanged. If cabling and equipment had to be re-installed each time a change was needed, it could be very expensive. The solution is to allow the users to design and administer their networks as they wish, without having to rearrange hardware. This is called a “Virtual” LAN, or VLAN. A VLAN is a network in the sense that - the number of users - the range of other networks and hosts that they can use are all carefully controlled, and the broadcast traffic is limited to that VLAN. A VLAN is not a physical network, but exists within a larger physical “box”, specifically a Virtual Network Hub. The same physical box may contain several VLANs.
In the diagram above, a VLAN switch has nine computers connected to it. Within the switch, they are organised into three separate VLANs, marked “A”, “B” and “C”. VLAN “A” has four computers attached to it, VLAN “B” has three, and VLAN “C” has two. Each of the VLANs appears to be a different network. For example, if the first computer on VLAN “A” sends out a broadcast, it only goes to the other three computers on VLAN “A”. It does not go to the other computers, even though they are connected to the same physical switch. Computers on VLAN “A” need special arrangements to talk to computers on VLAN “B”. This allows security to be controlled. (This special arrangement is called a router , which allows non-broadcast traffic to be routed to another network. Routers are described in more detail later, as they relate to Ethernet networks). At any time, the network manager can rearrange the VLANs. Using software alone, a computer can be “moved” from one VLAN to another, without changing any physical wiring or even moving a plug to a different socket. This gives the network manager a lot of freedom to optimise the organisation of the network, and control network security by limiting the scope of which computers are allowed to talk to each other.
While 10Mbps and 100Mbps systems are suitable for most users there is growing pressure on networks driven by factors that the original Ethernet designers couldnt have imagined. Traffic flows that traditioally terminated at a local resource are now increasing routed to the network core. User requirements hav changed significantly, 5 years ago word processing was the dominant application for PC users with a slightly smaller requirement for spreadsheet applications. Application useage has also chnaged with even basic word processors allowing the import of large data graphic files. Internet useage has accelerated the trend towards graphics even further.
Gigabit Ethernet is the logical development path for most users. All of the frame formats and protocols are identical allowing the user to increase speed without investment in new NOS systems. Management products such as SNMP management systems also remain unchanged. Gigabit Ethernet is primarily a fibre technology as the high speeds cause network distances to shrink in order to maintain the CSMA/CD architecture of Ethernet.
Gigabit Ethernet provides for 3 fibre types with SX being the most widely used (note that 1000BaseSX is in no way compatible with 100BaseSX). UTP Copper based products for 1000Base have yet to become available while the CX version using twinaxial cable is fairly expensive and very limited. It has so far been only available as a SANs connection system.
GBICs are the standard media carrier for Gigabit systems. These are generally provided as option ports to llow users to acquire additional fibre as required or change fibre modes as required. Unlike transceivers for 10 and 100Base systems Gigabit transceivers can be vendor proprietary for their connection to the switch product.
1000Base is mainly a switching product technology, however some very high end routing products also carry Gigabit interfaces. Hub products (because of the limited distances available using repeaters) have been aimed mainly at the storage area networks market where distance requirements are less of an issue. The two main types of Gigabit switch are Core devices which normally carry multiple Gigabit ports and Edge device which usually have limited Gigabit ports but multiple 10/100 ports. Core evices can be seen as a 'switch of switches' allowing stacks of switches to be built up using industry standards.
Gigabit Ethernet allows for very high performance t the network core. Most applications today are for network backbones to connect 10/100 devices to a higher speed regime for server access.
Ethernet Networking Technology Overview and Network Design
Agenda <ul><li>10 Mbit/s Ethernet technology </li></ul>What is a Local Area Network? Overview of network technologies 3 2 1
10 Mbit/s Ethernet technology What is a Local Area Network? Overview of network technologies 100 Mbit/s Ethernet technology Switching technology 1000 Mbit/s Ethernet technology 3 2 1 5 4 6
What is a Local Area Network? <ul><li>Definition of a LAN </li></ul><ul><li>Why use a LAN? </li></ul><ul><li>Characteristics of LANs </li></ul><ul><li>LAN Topologies and Structured Cabling </li></ul><ul><li>Repeaters, Bridges and Routers </li></ul><ul><li>Switching and Segmenting networks </li></ul><ul><li>Virtual LANs </li></ul>
Definition of a LAN <ul><li>A LAN is a system of cabling, equipment and software which allows computers to share and exchange data electronically, using an agreed format (protocol), within a ‘local’ area </li></ul>
Other types of Area Network <ul><li>WAN - Wide Area Network </li></ul><ul><ul><li>Interconnecting LANs and users over long distances, often on a public network </li></ul></ul><ul><li>MAN - Metropolitan Area Network </li></ul><ul><ul><li>Interconnecting LANs and users within a city area, typically by dedicated fibre optics </li></ul></ul>
Why use a LAN? <ul><li>An island of information </li></ul>
Why use a LAN? <ul><li>Users can share data </li></ul><ul><ul><li>Saves time, makes work more efficient </li></ul></ul><ul><li>Connect different computers together </li></ul><ul><ul><li>A LAN can be the common denominator </li></ul></ul><ul><li>Users can share resources (printers, storage) </li></ul><ul><ul><li>Saves money on expensive capital equipment </li></ul></ul><ul><ul><li>Centralised administration </li></ul></ul><ul><li>E-mail, Internet and Multimedia </li></ul><ul><ul><li>Reduces paper documents, better information </li></ul></ul>
LAN Characteristics <ul><li>How are LANs characterised? </li></ul><ul><ul><li>By speed (bandwidth) </li></ul></ul><ul><ul><li>By topology </li></ul></ul><ul><ul><li>By special features </li></ul></ul><ul><ul><li>By their target application </li></ul></ul><ul><li>LAN technologies are optimised for certain application areas </li></ul>
Network Topologies BUS Nodes are all connected to the same circuit RING Nodes are connected in a daisy chain
Network Access Methods Token Passing Nodes wait their turn to use the network Contention Nodes try to use the network at any time T
Network Wiring Topologies Bus Ring STAR wiring topology Nodes are physically wired to a central point
Structured Cabling <ul><li>Uses UTP for floor wiring </li></ul><ul><li>Uses Fibre for backbone connections </li></ul><ul><li>Many more wires installed than users - flood wiring </li></ul><ul><li>All cables star-wired from central points </li></ul><ul><li>Flexible, resilient, future proof, easy moves </li></ul>
Connecting nodes to a LAN Network Operating System (NOS) Software Network Interface Card (NIC) Cable Communication by an agreed Protocol Network Hub To other hosts on the network MAC Address “ DTE”
Network Protocols <ul><li>Protocols are the “language” used on a network </li></ul><ul><li>The network just sends and receives packets of data, while the protocol ensures that the right data is delivered to the right place </li></ul><ul><li>Using common protocols allows computers from different manufacturers to exchange data </li></ul>
Protocols defined... A protocol is a common system where both parties acknowledge the same rules governing communication.
Interconnecting Networks <ul><li>Three ways to join network segments together: Repeaters , Bridges and Routers </li></ul><ul><li>Repeaters extend physical networks </li></ul><ul><li>Bridges link networks of the same type together </li></ul><ul><li>Routers connect networks of different types together </li></ul>
Connecting network segments with Repeaters All network traffic visible to all the nodes on the network Repeater forwards all activity Repeater
Connecting network segments with Bridges Local Traffic stays on local segment Local Traffic stays on local segment Bridge only forwards traffic intended for the other network Bridge
Connecting networks together with Routers Router Router Local Traffic stays on local segments Routers transfer protocol-specific traffic (e.g. IP, IPX) between different network types, e.g. across a WAN X.25, Frame Relay, ISDN etc. Local Network Remote Network WAN
Summary <ul><li>Networks allow computers to share data quickly and cheaply </li></ul><ul><li>Networks are a combination of hardware and software </li></ul><ul><li>Network technologies can be shared or switched, or a mix of both </li></ul>
10 Mbit/s Ethernet technology What is a Local Area Network? 100 Mbit/s Ethernet technology Switching technology 1000 Mbit/s Ethernet technology Overview of network technologies 3 5 4 6 2 1
Ethernet Technologies <ul><li>10 Mbit/s Ethernet </li></ul><ul><ul><li>One of the oldest network technologies, and still the most popular </li></ul></ul><ul><li>Fast Ethernet (100 Mbit/s) </li></ul><ul><ul><li>Upgrade route from 10 Mbit/s, providing higher performance </li></ul></ul><ul><li>Gigabit Ethernet (1000 Mbit/s) </li></ul><ul><ul><li>The next generation for servers and backbones, providing very high throughput </li></ul></ul>
Gigabit Ethernet <ul><li>Promoted by the Gigabit Ethernet Alliance </li></ul><ul><li>Being standardised in IEEE 802.3z, due in 1998 </li></ul><ul><li>Uses same CMSA/CD technology as Ethernet, running at 1000 Mbit/s </li></ul><ul><li>Gigabit Ethernet on copper cabling is a special problem </li></ul><ul><li>High-performance backbone technology </li></ul>
Network Technologies: Summary <ul><li>Networking technologies can be classified by their access mechanism </li></ul><ul><li>Contention-based technologies are less efficient than other solutions, but the overall cost of ownership is lower </li></ul><ul><li>Ethernet covers all parts of a LAN from desktop to backbone </li></ul>
10 Mbit/s Ethernet technology What is a Local Area Network? Overview of network technologies 100 Mbit/s Ethernet technology Switching technology 1000 Mbit/s Ethernet technology 3 2 1 4 5 6
10 Mbit/s Ethernet <ul><li>Origins of 10 Mbit/s Ethernet </li></ul><ul><li>How Ethernet works </li></ul><ul><ul><li>CSMA/CD operation </li></ul></ul><ul><ul><li>Full and Half Duplex modes </li></ul></ul><ul><ul><li>Ethernet frames </li></ul></ul><ul><li>Components of an Ethernet LAN </li></ul><ul><ul><li>Software and Network Interface Card </li></ul></ul><ul><ul><li>Transceivers and Cabling </li></ul></ul><ul><ul><li>Repeaters and Hubs </li></ul></ul><ul><ul><li>Switches </li></ul></ul>
Origins of 10 Mbit/s Ethernet <ul><li>Original system design by DEC, Intel, and Xerox (hence DIX Ethernet) </li></ul><ul><li>Designed in 1970’s, first specifications 1980 </li></ul><ul><li>Ethernet type II adopted as IEEE 802.3 10BASE-X, first published in 1985 </li></ul>
How Ethernet works 1 <ul><li>All transmission is at 10 million bits per second (0’s or 1’s) </li></ul><ul><li>Users are connected to common cable (media) </li></ul><ul><li>Access to the media is by a simple set of rules known as Carrier Sense Multiple Access / Collision Detect (CSMA/CD). </li></ul><ul><ul><li>Listen for silence on cable (CS) </li></ul></ul><ul><ul><li>Transmit data without waiting your turn (MA) </li></ul></ul><ul><ul><li>If you hear someone else talking - stop sending, and wait for a random time before trying again (CD) </li></ul></ul>
How Ethernet Works 2 1. Send when the network is quiet 2. Collision is detected if another station sends 3. Both stations wait for a random time 4. Re-send again when the network is quiet
Full and Half Duplex 1 <ul><li>Normal Ethernet only allows one frame on the cable at a time (Half Duplex) </li></ul><ul><li>UTP and fibre optics use separate circuits for Transmit and Receive </li></ul><ul><li>Full Duplex allows frames to be sent and received at the same time over a point-to-point link </li></ul><ul><li>Both ends must support Full Duplex </li></ul><ul><li>Repeaters cannot support full duplex </li></ul>
Full and Half Duplex 2 Half Duplex Operation Transmit only - OK Receive only - OK Transmit and Receive = Collision Full Duplex Operation Transmit and Receive at same time on separate circuits - OK
How Data is Transferred <ul><li>All data is transferred in ‘packets’ </li></ul><ul><li>A packet of data has addressing details at the start, and error checking data at the end. This is known as a ‘ frame ’ </li></ul><ul><li>Moving data in small pieces gives everyone an equal chance to get their data through </li></ul><ul><li>Smaller packets are more likely to be delivered without errors </li></ul>
The Ethernet Frame <ul><li>Preamble allows timing alignment </li></ul><ul><li>Start Of Frame delimiter indicates start of frame </li></ul><ul><li>CRC (Cyclic Redundancy Check) is a checksum to ensure the frame was received OK </li></ul><ul><li>Total frame length varies from 64 to 1,518 bytes (after SOF delimiter) </li></ul>Preamble CRC (checksum) Data 46 - 1500 Bytes Source MAC Address Destination MAC Address SOF delimiter Type/length 6 bytes 6 Bytes 2 4
Components of an Ethernet LAN Shared cable External Transceiver Drop cable (external XCVR) Network Interface card Software Point-to-point link cable (integral XCVR) Hub: Repeater, Bridge, Switch or Router
Network Interface Card <ul><li>The Network Interface Card (NIC) contains: </li></ul><ul><ul><li>the connection to the transceiver, or a built-in transceiver </li></ul></ul><ul><ul><li>circuitry for generating frames and accessing the network </li></ul></ul><ul><ul><li>the physical MAC address </li></ul></ul><ul><ul><li>a software interface to the protocol software in the host </li></ul></ul>
Transceivers <ul><li>Transceivers provide the electrical and physical connection between the Adapter and the shared network cable </li></ul>XCVR <ul><li>This type of transceiver is not used much today as most NICs have this function built in. </li></ul>AUI drop cable Thick Ethernet Cable
Current Transceiver Uses <ul><li>Connecting standard hub/router/switch AUI interfaces to cable media </li></ul><ul><li>Connecting ‘legacy’ cards to newer cabling </li></ul>
Ethernet Cable Options <ul><li>The cable provides physical connection between the adapter cards. Multiple cable types are available. </li></ul>10Base5 10Base2 10BaseT 100BaseTX 10BaseFL 100BaseFX Thick Ethernet Thin Ethernet Twisted Pair (UTP) Fibre Optic BNC RJ45 SMA* Screw type ST Bayonet SC Dual Mini AUI connection via Vampire Tap Ethernet Cable Connector Usage Half-duplex shared cable Full-Duplex Point-to-point link Half-duplex shared cable Full-Duplex Point-to-point link * SMA now obsolete, no new equipment manufactured to support this standard
Ethernet Cable Options 10BaseFL 100BaseFX Fibre Optic Lucent LC New sub-miniature SC 3M Volition VF-45 Fibre version of RJ-45 AMP MT-RJ Fibre version of RJ-45 New fibre connectors becoming widespread during 1999/2000. Full-Duplex Point-to-point link
10 Mbit/s Repeaters <ul><li>10 Mbit/s Repeaters allow more users AND more distance </li></ul><ul><li>Repeaters do a number of tasks: </li></ul><ul><ul><li>restore the signal levels (amplify signal) </li></ul></ul><ul><ul><li>ensure that collisions are recognised, and stop anyone else transmitting until it is safe </li></ul></ul>
<ul><li>Si può vedere l’hub come un multiport repeater. </li></ul>Four Repeater Roule <ul><li>Come nel coassiale esiste la regola dei 4 repeater negli hub a 10Mb/s </li></ul>Hub Hub Hub Hub
Ethernet 5-4-3 Rule <ul><li>Maximum of five segments </li></ul><ul><li>Maximum of four repeaters between any two nodes </li></ul><ul><li>Maximum of three multi-node segments </li></ul>Repeater Repeater Repeater Repeater Multinode Segment Multinode Segment Multinode Segment Link Segment Link Segment
Why are there Limits? <ul><li>Transmission from a user down the network cable takes time . </li></ul><ul><li>All users must see transmission before user has transmitted half of his frame. </li></ul><ul><li>This may need to checked by calculating the Round Trip Delay Time . </li></ul><ul><ul><li>i.e. RT Delay < time for min frame length </li></ul></ul><ul><ul><li>or time for 64 Bytes (512 Bit times) </li></ul></ul><ul><li>The 5-4-3 rule is usually sufficient for 10 Mbit/s Ethernet </li></ul>
Ethernet Hubs <ul><li>Hubs provide a central connection point for networks </li></ul><ul><li>Commonly used with Structured Cabling Schemes </li></ul>
Hubs and Management <ul><li>Hubs can be combined to create one big repeater - stackable, chassis based </li></ul><ul><li>Flexible options for integrating all media types </li></ul><ul><li>Management features allow network supervisors to see traffic flow and solve problems fast </li></ul><ul><li>Other advanced features also added </li></ul><ul><ul><li>e.g. security </li></ul></ul>
10 Mbit/s Ethernet: Summary <ul><li>Repeater </li></ul><ul><ul><li>copies everything from one segment to another: collisions, fragments, all frames including broadcasts </li></ul></ul><ul><li>Bridges and Switching Hubs </li></ul><ul><ul><li>Selected frames including broadcasts are copied, based on the destination MAC address </li></ul></ul><ul><li>Router </li></ul><ul><ul><li>Copies / converts selected frames based on protocol address </li></ul></ul>
10 Mbit/s Ethernet technology What is a Local Area Network? Overview of network technologies 100 Mbit/s Ethernet technology Switching technology 1000 Mbit/s Ethernet technology 3 2 1 4 5 6
100 Mbit/s Ethernet <ul><li>Relationship with 10BASE-X </li></ul><ul><li>Media options and technologies </li></ul><ul><li>Repeaters and configuration rules </li></ul><ul><li>Media Independent Interface (MII) </li></ul><ul><li>N-Way Auto-Negotiation </li></ul><ul><li>100BASE-X switching </li></ul>
100BASE-X Networks <ul><li>Based on CSMA/CD </li></ul><ul><li>Transmission at 100 Million Bits per second </li></ul><ul><li>Uses same frames as 10 Mb Ethernet </li></ul><ul><ul><li>Whole protocol stack and NOS remain unchanged </li></ul></ul><ul><ul><li>Easy migration from existing systems </li></ul></ul>
Auto-Negotiation (Nway) <ul><li>Nway auto-negotiation </li></ul><ul><ul><li>Negotiation of a ‘way’ from N options </li></ul></ul><ul><ul><li>Happens between the two ends of a link </li></ul></ul><ul><ul><li>Fastest available ‘way’ is selected </li></ul></ul><ul><ul><li>If only one end has Nway then speed is detected (but not always duplex mode) </li></ul></ul>
MII - Media Independent Interface <ul><li>MII is a standardised interface between a 100BASE-X transceiver and the connected station (DTE) </li></ul><ul><li>The DTE can control the transceiver function, e.g. full/half duplex & speed </li></ul><ul><li>The transceiver can ‘declare’ its capabilities to the DTE </li></ul>
100BASE-X Media Options <ul><li>Cable I EEE 802.3u Standard Pairs Used Distance </li></ul><ul><li>UTP Cat 5 100BASE-TX 2 100m </li></ul><ul><li>UTP Cat 3,4,5 100BASE-T4 4 100m </li></ul><ul><li>Fibre 100BASE-FX 2 fibres 412m* </li></ul><ul><li>Fibre 100BASE-FX 2 fibres 2000 m* * </li></ul><ul><li>* Half Duplex DTE-DTE </li></ul><ul><li>* * Full Duplex DTE-DTE </li></ul>
100BASE-X Technologies: TX <ul><li>100BASE-TX </li></ul><ul><ul><li>Uses two twisted pairs </li></ul></ul><ul><ul><li>Same system as 10BASE-T but faster! </li></ul></ul><ul><ul><li>Must have Category 5 cable to run on </li></ul></ul><ul><ul><li>Max distance 100m </li></ul></ul><ul><ul><li>Full or Half duplex possible </li></ul></ul>
100BASE-X Technologies: FX <ul><li>100BASE-FX </li></ul><ul><ul><li>Uses two 62.5/125 multimode fibres </li></ul></ul><ul><ul><li>Operates at 1300nm (10BASE-FL operates at 850nm) </li></ul></ul><ul><ul><li>Full duplex is possible but no auto-negotiation </li></ul></ul><ul><ul><li>The maximum link length depends on the configuration of the network </li></ul></ul><ul><ul><ul><li>i.e. it gets shorter if you have a repeater </li></ul></ul></ul>
100BASE-X Repeaters <ul><li>Primary use for 100BASE-X repeaters is to add more users </li></ul><ul><li>Otherwise the same functions as 10BASE-X repeater </li></ul>
100BASE-X Repeater Classes <ul><li>IEEE 802.3u defines two classes of repeater according to signal delay </li></ul><ul><ul><li>Class I (the worst class!) </li></ul></ul><ul><ul><ul><li>delay of less than 140 bit times </li></ul></ul></ul><ul><ul><li>Class II </li></ul></ul><ul><ul><ul><li>delay of less than 92 bit times </li></ul></ul></ul><ul><li>1 bit time is 1/100,000,000th of a second (10 nanoseconds) </li></ul>
Basic Configuration Rules <ul><li>Maximum of one Class I repeater in a collision domain </li></ul><ul><li>Maximum of 2 Class II repeaters, but limited distance </li></ul>100m Class I 100m 100m Class II 100m 5m Class II
Calculating complex configurations <ul><li>As with 10 Mbit/s Ethernet, the Round Trip Delay time must be less than 512 bit times to guarantee collision detection </li></ul><ul><li>The basic rules about repeaters onl y cover the situations in the previous slide, so always calculate anything else </li></ul>
Calculating the Delay <ul><li>“ Typical bit time delays ” for 100 Mbit/s network components </li></ul>Component Delay per metre Max delay Two TX/FX DTEs 100 Two T4 DTEs 138 One T4 and one TX/FX DTE 127 Cat 3 Cable Segment 1.14 114 (100m) Cat 4 Cable segment 1.14 114 (100m) Cat 5 Cable segment 1.112 111.2 (100m) Shielded Twisted Pair Cable 1.112 111.2 (100m) Fibre Optic Cable 1.0 412 (412m) Class I Repeater 140 Class II Repeater all TX/FX 92 Class II Repeater with any T4 67 AT-MC101 40
Example Calculation 1 100Mbps Hub 60m Fibre 80m UTP 100m UTP Device Delay (bit times) Two DTEs 100 180m Cat 5 UTP 200 2* Class II Repeaters 184 60m Fibre segment 60 margin 4 TOTAL 548 The above system is outside the limits and will not function correctly 100Mbps Hub
Example Calculation 2 200m Fibre 100m UTP Device Delay (bit times) Two DTEs 100 100m Cat 5 UTP 111 Class II Repeater 92 200m Fibre segment 200 Margin 4 TOTAL 507 The above system is within the 512 bit time limit so will work OK. 100Mbps Hub
Switched 100 Mbit/s <ul><li>Similar to switched 10 Mbit/s Ethernet </li></ul><ul><li>Most 100 Mbit/s switches support 10 Mbit/s ports too </li></ul><ul><li>Requires an order of magnitude increase in throughput </li></ul><ul><ul><li>160 Mb/s for 16 port 10Mbps switch </li></ul></ul><ul><ul><li>1.6 Gb/s for 16 port 100Mbps switch </li></ul></ul><ul><li>Key application as backbone switch </li></ul>
100 Mbit/s Ethernet: Summary <ul><li>100 Mbit/s and 10 Mbit/s Ethernet use the same CSMA/CD technology and software </li></ul><ul><li>Shared multi-drop cables are not supported </li></ul><ul><li>Round Trip Delay time calculations are essential for network design </li></ul><ul><li>100 Mbit/s can be integrated into a 10 Mbit/s network only with switching technology </li></ul>
100 Mbit/s Ethernet: Summary <ul><li>Distances are more restricted so 100Mbps tends to be switch-centric in design. </li></ul>H H H H H H H H H H H S S S S S S S
10 Mbit/s Ethernet technology What is a Local Area Network? Overview of network technologies 100 Mbit/s Ethernet technology Switching technology 1000 Mbit/s Ethernet technology 3 2 1 5 4 6
Switching Technology 1 <ul><ul><li>The maximum bandwidth available on a shared network is limited to the network speed, regardless of the number of nodes </li></ul></ul><ul><ul><li>The maximum traffic that these nodes could generate is more than the network could carry </li></ul></ul><ul><ul><li>Sharing the same media introduces a bottleneck </li></ul></ul>Conversation
Switching Technology 2 Three simultaneous conversations <ul><ul><li>A Switching Hub at the centre of the network can handle multiple point-to-point conversations at the same time </li></ul></ul><ul><ul><li>The use of the network is more efficient </li></ul></ul>
Segmenting Networks 1 <ul><li>Very large networks are less efficient, even with switching technology </li></ul><ul><li>It is unusual for all users on a large network to need frequent communications with all other users </li></ul><ul><li>Segmenting a large network into functional groups improves performance and manageability </li></ul>
Segmenting Networks 2 100% <ul><li>Large networks generate a lot of traffic at the hub in the centre </li></ul><ul><li>The broadcast load is proportional to the square of the number of nodes </li></ul>
Switched 10 Mbit/s <ul><li>Switch separates traffic that is sent direct from station to station - Datagrams . </li></ul><ul><li>MAC addresses are learned from incoming frames. </li></ul><ul><li>Frames are only sent out of port where the destination address resides. </li></ul><ul><li>Broadcast frames are forwarded to all ports, to the MAC address FF-FF-FF-FF-FF-FF </li></ul>
Switching Techniques 1 <ul><li>Store and forward switching </li></ul><ul><ul><li>Whole frame buffered in memory, then sent </li></ul></ul>Preamble CRC Data 46 - 1500 Bytes Source Address Destination Address SOF delimiter Type/length 6 bytes 6 Bytes 2 4 Cut-through switching Forwarding started just after the destination MAC address arrives
Segment Switching <ul><li>Switch is used to ‘feed’ hubs. </li></ul><ul><li>Small groups of users share 10 Mbit/s segments from switch e.g. Turbo Stack </li></ul><ul><li>Reasonable MAC address capacity required </li></ul><ul><li>Extra features </li></ul>hub hub hub hub switch
Workgroup Switching <ul><li>Small switch provides dedicated switch ports to individual ‘power’ users. </li></ul><ul><li>Low MAC address capacity required </li></ul><ul><li>Low cost per port </li></ul>Switch
Backbone/Enterprise Switching <ul><li>Large, high capacity switch used to connect backbone segments of large networks. </li></ul><ul><li>Huge MAC address capacity required </li></ul><ul><li>Modular design - chassis based </li></ul><ul><li>Very high throughput capability </li></ul>
Bandwidth Switching <ul><li>Switches can be used to connect segments of different speeds, e.g. 10 and 100 Mbit/s </li></ul><ul><li>The 10 Mbit/s segments can be ‘multiplexed’ onto the 100 Mbit/s segment </li></ul><ul><li>The switch must use Store-and-Forward to change the bandwidth </li></ul>Switch 100 Mbit/s 10 Mbit/s
Rules for using Switches <ul><li>Switch is essentially a bridge , so start any repeater counts again. </li></ul><ul><li>Too many switches in network could cause delay. IEE 802.3 says max 7 bridges. (This is an arbitrary number based on Token Ring limitations) </li></ul><ul><li>Keep number of hops to main servers low. </li></ul>
Hub Dual-Speed Autosensing <ul><li>Reti miste: con utenti a 10 e 100 mbps </li></ul><ul><li>Configurazione Stackable: facilmente espandibile </li></ul><ul><li>Percorso di migrazione da Ethernet a Fast Ethernet </li></ul>Sw 10 Mbps 100 Mbps … . Switch Module
Virtual LANs (VLANs) 1 <ul><li>A company-wide network may not be the right solution, due to high traffic and security issues </li></ul><ul><li>The user may want to divide their big network into smaller sections </li></ul><ul><li>But, flexibility is also needed as people move around </li></ul><ul><li>A “Virtual Network” solves this problem </li></ul>
10 Mbit/s Ethernet technology What is a Local Area Network? Overview of network technologies 100 Mbit/s Ethernet technology 1000 Mbit/s Ethernet technology Switching technology 3 2 1 6 4 5
The pressure on networks today <ul><li>Traffic Flows have changed </li></ul><ul><ul><li>Used to be 80% local, 20% backbone </li></ul></ul><ul><ul><li>Now 20% local and 80% backbone </li></ul></ul><ul><li>User requirements have changed </li></ul><ul><ul><li>Word processing was dominant </li></ul></ul><ul><ul><li>Now Internet use is the dominant application </li></ul></ul><ul><li>Application useage has changed </li></ul><ul><ul><li>It only takes 8 bytes to send the word ‘airplane’ </li></ul></ul><ul><ul><li>It takes 80,000 bytes to send an image of an airplane </li></ul></ul><ul><ul><li>It takes 8,000,000 to send a video clip of an airplane </li></ul></ul>
1000Mbps Gigabit Systems <ul><li>Same frame formats and protocols as 10/100 Ethernet </li></ul><ul><li>Same Full/Half Duplex mode </li></ul><ul><li>Same management (SNMP/RMON) systems </li></ul><ul><li>Primarily a fibre based technology </li></ul><ul><li>Standardised by IEEE as 802.3z </li></ul>
1000BaseX Cable Options <ul><li>Standard Media Range </li></ul><ul><li>1000BaseLX Single Mode Fibre 3km+ </li></ul><ul><li>1000BaseLX Multi-Mode Fibre 550m </li></ul><ul><li>1000BaseSX Multi-Mode 300m </li></ul><ul><li>(most common type) </li></ul><ul><li>1000BaseCX Twinaxial Cable 25m </li></ul><ul><li>1000BaseT UTP 100m </li></ul>
1000BaseX Media <ul><li>The basic media carrier for 100Base is the GBIC </li></ul><ul><li>GBICs are a media independent device (similar to transceiver in 10Base) </li></ul><ul><li>Provide plug in options for SX or LX fibre </li></ul>
1000BaseX Hardware <ul><li>Switches and Routers for Networks </li></ul><ul><li>Two primary types of switch </li></ul><ul><ul><li>Core Switches…… Multiple Gigabit ports </li></ul></ul><ul><ul><li>Edge Switches…… Multiple 10/100 ports 1 or 2 Gigabit ports </li></ul></ul>