 |
Home | Products | Services | Case Studies | Technical Information | Press Room |
 |
|
|||||
2. MAIN TYPES OF LANThe operation of a LAN can usually be separated into two main aspects. Firstly – the physical medium (connector types, voltage and electrical signals) and the method of placing data onto the network. In OSI systems this corresponds to layer 1 and the lower part of layer 2 of the reference model. Secondly – the operating software which establishes end-to-end transmission with guaranteed data delivery between two devices, communicating across the network. In OSI systems this corresponds to the upper part of layer 2, layer 3 and layer 4 of the reference model. Carrier Sense Multiple Access with Collision Detect (CSMA/CD) – Ethernet‘Ethernet’ is one of the most widely known terms in LAN technology. The term derives from the original network which was defined by Xerox and adopted by several other organisations including DEC (Digital Equipment Corporation) and Intel. The original published specifications were known as DIX (Dec, Intel and Xerox) Ethernet Specifications Versions 1 and 2. The Institute of Electrical and Electronic Engineers (IEEE) adopted, improved and modified the DIX version 2 specification and this has become the IEEE 802.3 standard, which equates to the ISO 8802/3 standard. Carrier Sense Multiple Access with Collision Detect (CSMA/CD) networks operate using a bus structure, that is a single strand of cable to which all devices connect. It uses baseband communications (i.e. only one signal can travel on the cable at one time) at 10Mbps (10,000,000) bits per second, although the original 10Mbps gone on to become 100Mbps and now Gbps. Most Servers and Ethernet switches support links of 1Gbps, with higher end devices supporting 10Gbps with 100Gbps expected soon. Carrier Sense Multiple Access with Collision DetectionUsing Carrier Sense Multiple Access with Collision Detection means all devices on the LAN are free to communicate whenever they need to without any precedence or order. A device wishing to send monitors the network (Carrier Sense) and, if no other device is sending, begins to transmit. It is possible that another device will also start to transmit at that moment (Multiple Access), so the device checks for a collision (Collision Detect). If a collision occurs, (this is the transmitting station detects another station on the LAN) then all devices involved in the collision stop, the device that was transmitting the frame, transmits a jam signal, and pauses for a period, of time known as the ‘back off delay’ (which is determined using the truncated binary exponential backoff algorithm) before trying to send that frame again. All devices monitor the network continuously, copying and acknowledging all packets addressed to that device. Accessing a network in this way is known as probabilistic or non-deterministic. Probabilistic because the ability of any one station to transmit on the network is based on the level of activity on the network: The higher the level of activity the lower the chance. Non-deterministic because the designer is unable to guarantee the level of performance or delay which will be experienced by any one station on the network under particular loading conditions. The original 10 Base 5 (Thick Ethernet) topology of Ethernet is that of a branching tree structure with interconnecting segments (see figure 1). A loop in the interconnection segments must be avoided. Each segment can be up to 500 metres in length with a maximum number of 100 network nodes (or taps) per segment. To extend beyond the maximum length or number of devices, segments are linked together with repeaters or half repeaters. These simply extend the length of the network by effectively regenerating and repeating the signal. A repeater connects two local network segments. A half-repeater implements a transmission line between two segments therefore enabling a greater distance to be spanned between two segments. There is a limit of four repeaters or half repeaters, which can be supported between any two points on the network. To extend the network further bridges and routers can be used.
Figure 1: CSMA/CD Tree Structure A significant factor of original Ethernet or CSMA/CD networks was the cost of the co-axial cable. The IEEE standard specifies the use of quality coaxial cable (“Yellow cable”) or a thinner cheaper co-axial cable (RG58 specification). These are termed 10 Base 5 and 10 base 2, referring to 10 Megabits per second (MBPS) BASEband transmission 500 metres maximum segment length, and 10 Mbps BASEband transmission 200 metres maximum segment length respectively. 10 Base 2 is often referred to as “Cheapernet” and has an actual maximum segment length of 185 metres and a maximum of 30 taps per segment. At the time of writing this booklet 10 Base 5 and 10 Base 2 are less common and it is far more common to see 10BASE T (10/100 Mbps BASEBand transmission on twisted pair cable). This has a maximum segment length of 100 metres and 1 tap, i.e. it is point-to-point only. Ethernet is the most common form of LAN technology installed today due to its early arrival in the marketplace. Originally there were two important concerns regarding this type of technology. Firstly – the term Ethernet does not guarantee compatibility of hardware, as there are three different standards:
Previously it was imperative to see confirmation before installing some of the older DIX network components on IEEE type networks. However it would be unusual to purchase any non-802.3 devices today, so the issue of incompatibility is less likely to arise nowadays. Secondly – degradation of performance under loading is non-linear and the performance of most networks can degrade significantly under sustained heavy network loads. Because the term ‘Ethernet’ was in generic usage, it was important that specifiers defined exactly what level of standards compliance was required (e.g. IEEE 802.3/ISO 8802-3). Today it would be unusual to purchase products which do not conform to IEEE 802.3. CSMA/CD EfficiencyCSMA/CD has a minimum frame size of 64 bytes, (if the payload is smaller the network will pad the frame out to 64 bytes) in order for the collision detection mechanism to work, and a maximum frame size of 1518 bytes (1522 bytes when running Tagged VLANs). Token RingToken Ring is based on a closed loop philosophy so that eventually a station will receive its own transmission. The token is a single special sequence, which circulates around the loop, with each station on the ring receiving and regenerating the token. When a station wishes to transmit data, it waits for the token, adds addressing information plus the data, marks the token busy and then sends the token to the next station. All the stations on the network continue to receive and regenerate frames, but if a station wishes to send data it must wait for the token to become free. A station, which receives a token addressed to itself, copies the data and regenerates the frame. Eventually the sending station detects the return of its busy frame, removes it and then transmits a free token, giving the next station an opportunity to send.
Figure 2: IBM Token Ring Structure A new form of token passing was developed to improve network efficiency. It was called ‘early token release’ and allowed the token to be released immediately after a data frame had been transmitted. This reduced the delay time, as the station no longer had to wait for its data packet to return, which could take a considerable time on a network with many devices. There are many types of network employing Token Ring – The IEEE 802.5 standard. The most common is the IBM Token Ring System originally operating at 4Mbps, and then 16Mbps. The standard specification uses a twisted pair cable running baseband communication at 4Mbps. This offers a cost advantage over the original Ethernet, which operated on co-ax cable, as twisted pair cabling is cheaper. However the IBM full specification screened cabling can be expensive. The introduction of 10Base T Standards for Ethernet (Ethernet operating on twisted pair cable) has effectively addresses this difference by allowing transmission over unscreened cable. While Token Ring is the architecture of this type of LAN, the IBM Token Ring network need not be a physical ring topology (see figure 2). A device called a Multi-Station Access Unit (MAU) will act as the centre of a star-based ring topology. Unlike Ethernet, Token Ring is not naturally resilient and the removal of a station in the ring would cause all data to stop. To protect against this, the MAU monitors each attached device and heals the ring should a break occur. A MAU supports a number of attached devices (usually seven) and then attaches to other MAUs in the network. In fact the MAU may support one or more sub rings on any of its connections rather than a single device. As with CSMA/CD network repeaters can be used to extend the ring, possibly between two buildings, although repeaters do not increase the maximum number of devices, which can be supported by the network. As previously stated, the main limitation of a ring topology is that a break in the ring causes the whole network to fail. The MAU maintains an active configuration path so that any failure is detected and circumvented immediately, causing the ring to recover gracefully, so that all users, except those on the failed section, will be unaware that the failure occurred. One of the consequences of this type of fault is that a device could fail or become disconnected while it still has the token, and the token may become lost, or a device may fail after transmitting a busy token and therefore be unable to release the token. In both cases an arbiter is responsible for detecting the anomalous condition and taking corrective action. Where MAUs are not used, a device is available for connection between the ring and attached device, which produces the same recovery function as the MAU. The main limitation on the topology is a maximum distance of 100 metres between stations (this allows for the failure of a station at 200 metres, which can be supported during failure). A ring supports a maximum of 33 MAUs and 260 stations, although the network can be extended past these limitations by using bridges and routers to link the rings. The benefit of Token Ring is that a station can only hold the token for a predetermined period, thus giving all stations an opportunity to transmit on a regular basis whatever the level of traffic on the network. Another benefit over CSMA/CD or Ethernet is that there are no collisions and therefore the performance degrades linearly under heavy loading. Token BusToken Bus combines the bus structure of Ethernet type networks and the token system in Token Ring. The standard form of transmission uses broadband communication on co-axial cable. Broadband communication divides the signals on the network into different frequencies, allowing more than one signal to travel on the cable at any one time. This can be compared with the use of co-axial cable for carrying several television signals simultaneously. Signals are normally generated in pairs and one cable may support several different pairs. A variety of speeds may be used, four 1 Mbps pairs, one 5 Mbps pair or one 10Mbps pair. The most common form currently is the 5 Mbps pair. The network needs to be able to transmit to all devices on the bus. Therefore the signal is divided, and two channels – forward and reverse – are implemented. When a signal reached the head end of the network the signal on one channel is re-modulated (i.e. the frequency is changed) and then output on the other channel. This allows any station to broadcast to any other station, regardless of its position on the network. As token passing is implemented, and the network does not form a ring, a logical ring is implemented. This uses the addresses of the devices on the network and each device transmits the token to the next logical address on the bus. The use of broadband requires a more complicated signalling system and involves a form of modem for each device attached to the network. The network also requires a device at the head end to re-modulate and regenerate the signals. It can therefore be more expensive to implement than baseband. The specification is covered by the IEEE 802.4 standard. While Token Bus is not widely used, the most common implementation is in Manufacturing Automation Protocol (MAP) networks. The benefits of Token Bus are that cabling is far easier to implement than a ring topology and superior performance to CSMA/CD can be achieved under high loading conditions by collision avoidance, as token passing is implemented. However, as a logical loop is employed, the token must be captured and regenerated before it can be sent to the next device in the logical loop, which produces large overhead on the network. To reduce this overhead, multiple transmissions can be implemented during token capture by the device seizing the token. However, this only partially resolves the problem as the token can only be held for a limited period. Fibre Distributed Data Interface (FDDI)FDDI is a standard issued by the American National Standards Institute (ANSI). It is based on Fibre optic cable, token passing access methods and a ring topology. It is effectively a token ring network, which can be up to 100km in length, and operates at 120Mbps, but after removing overheads, provides useable bandwidth of 100Mbps. With a maximum distance of 100km this network belies the term local area network. The network should perhaps be regarded as a backbone, linking buildings and central resources to a series of small, lower cost LANS in each department or floor as required. Such reasonable capacity, long distance, backbone networks are often referred to as Metropolitan Area Networks (MANs). Another application area for FDDI could be for, more specialised workstations such as those used in Computer Aided Design (CAD) where large amounts of data may need to be transferred from host computer to terminals on a frequent basis (see figure 3 ‘FDDI Structure’).
Figure 3: FDDI Structure FDDI offers several key benefits over conventional networks. Firstly – the specifications implements dual counter rotating optical rings bestowing fault tolerance to the ring and attached nodes. Secondly – more than one packet can travel the network at the same time, allowing better use of the large size of the network and capability of the optical cable. Thirdly – the maximum packet size is much larger than other networks, thus enabling efficient data transfer, especially for devices using particularly large amounts of data, such as graphics workstations. Fourthly – since fibre optic cable uses light it is free from all normal forms of electrical interference. Errors in the data are therefore very low and few retransmissions are required, increasing the bandwidth available for attached devices. Finally – the use of token passing eliminates collision problems and this has been further developed to allow different devices to be prioritised for network usage. Therefore, key devices and those with high data volumes can be given priority, eliminating possible delays. The four standards for FDDI are:
Other types of LANMany proprietary networks were developed the most common of which was Apple Talk and ARCnet. Some proprietary networks offered ‘standards conformance’ but this phrase may be misleading as they may have been developed using a model similar to the OSI model, but not one, which is fully conformant. The only guarantee of standards conformance is certification by an independent testing body.
|
||||||
| < Previous section | Next section > |
 Top |
Copyright © Case Communications 2005 |  |