Categorizing Networks by Architecture
Generally speaking the network architecture is highly related to the
physical and logical topology. They are usually related to the types
of protocols used on the LAN in addition to the topology. The most
popular LAN architectures are
Though Ethernet is decidedly the most prevalent and almost completely displacing token rings.
Ethernet. Ethernet was first developed at Xerox in the 1960s. After this a consortium was formed along with Digital Equipment Company and Intel to create the DIX Ethernet. This has essentially led to the IEEE 802.3 specifications.
Wired Ethernet networks are usually configured as a physical bus or star and use the CSMA/CD - Carrier Sense Multiple Access Collision Detection method of media access. We leave wireless for later pages. Devices sharing an Ethernet listen for other traffic and transmit only if the LAN is clear. If two stations send at about the same time and their messages collide, both transmissions are aborted and they wait for a randomly generated period of time before retransmitting. Unfortunately, this sort of collision mechanism results in quite a bit of inefficiency on the network. But on the other hand they Ethernets are usually much less expensive than their competition.
Ethernets typically come in three main varieties based on their capacity. The standard is 10 Mbps, Fast Ethernet is 100 Mbps. Gigabit Ethernet is starting to be used more and more and is capable of carrying data at speeds in excess of 1 Gbps.
There are different subcategories of Ethernet mostly based on the cabling that is used. The following list outlines some of the most prevalent.
The following diagram illustrates the meaning of the short notations used in the preceding list.
|The following sections survey characteristics of
Ethernet networks based each of these types of cabling.
Coax Based Ethernets
10Base5 Networks. These Ethernets use thick coax cable also called thicknet networks. The 10 in 10Base5 derives to 10 Mbps speed. The 5 derives from the maximum length which is 500 meters.
One reason this cabling can be somewhat more difficult to work with than other connective devices because it is less flexible. It also the case that connections require what is called a vampire tap where a small hole is drilled into the cable to attach it to each connector. 10Base5 networks also make use of external transceivers to generate and receive the data signals. In almost all other Ethernet configurations the transceivers are built into the device NICs. 10Base5 makes use of DIX connectors with AUI cable to connect the transceiver to the NIC.
10Base2 Networks. This is another popular coax based network type. This uses thinner, less expensive and more flexible coax than the 10Base5. The 10 again refers to the 10 Mbps speed. The 2 is actually an upward round off that refers to the upper limit of 185 meters in length.
These are also somewhat easier to setup than 10Base5. For example, these networks make use of twist and push connectors, called BNC connectors, that are then connected to a T-connector on the network interface card. The transceivers are built into the network cards. The cable is called RG-58A/U or RG-58C/U.
UTP Ethernet Networks
The most popular cabling for Ethernets is UTP - Unshielded Twisted Pair. While there are many variants of UTP, most everyone is familiar with it because it is used for telephone cabling. UTP comes in different grades numbered as categories and summarized in the following table.
|UTP Category||Maximum Transmission Speed||Characteristics and Uses|
|Cat 1||Voice only||Used in telephone installations.|
|Cat 2||4 Mbps||No longer recommended for data transmissions.|
|Cat 3||16 Mbps||Lowest recognized data grade. It is used for most telephone wiring.|
|Cat 4||20 Mbps||Suitable for networking 10 Mbps Ethernet networks.|
|Cat 5||100 Mbps - 1Gbps||Most popular grade for LAN networking. This is used for Fast Ethernet.|
|Cat 5e (enhanced)||155 Mbps||Used for Fast Ethernet and 155 Mbps ATM - Asynchronous Transfer Mode.|
|Cat 6 and 7||> 1 Gbps||Used for Gigabit Ethernet technologies.|
|Notice that almost all of these support faster
transmission speeds than coax. UTP is also highly flexible and,
relative to coax, easy to install. It uses RJ connectors, which are
the modular plug types used for telephones. Most telephone cables
use an RJ-11 connector. Most Ethernet cables connect with a slightly
larger RJ-45 connector.
10BaseT Networks. This is popular for LANs of all sizes. It can run on Cat 3 cable which is already installed in most buildings for telephone communications. Newer 10BaseT networks are usually set up using category 5 cabling so that it is easy to upgrade to 100 Mbps later.
100BaseT Networks. 100BaseT refers to Ethernet networks running at 100 Mbps over Cat 5 or 5e cabling. Networks operating at these speeds are more often called Fast Ethernet. These networks use the same topology and access methods as 10BaseT. In fact, the only real differences are the they require a faster network card and hubs or switches that support the 100 Mbps transmission speed.
Many hubs/switches and NICs support both 10 Mbps and 100 Mbps transmission speeds making it much easier to upgrade later. You can even run some LAN segments at 10 Mbps and others at 100 Mbps as you upgrade.
1000BaseT Networks. This is more commonly called Gigabit Ethernet. It was established in 1996 as the IEEE 802.3x specifications. These provide for the speed, the Ethernet frame format and the CSMA/CD transmission method.
Fortunately, Gigabit Ethernet has been developed to be interoperable and backward compatible with its predecessors. It can make an excellent backbone technology with 10BaseT and 100BaseT LANs. Unfortunately, NICs, hubs and switches that support Gigabit Ethernet are several times more expensive than their predecessors.
100BaseVG-AnyLAN Networks. Hewlett Packard developed the 100BaseVG-AnyLAN for 100 Mbps Ethernet that makes use of intelligent hubs for central control. This hub manages network access by constantly doing a sequence of scans of all network port requests to check for service requests. The hub receives the incoming data packet and directs it only to the port with a matching destination address. This matching improves both security and reduces broadcasting on Ethernets to help decrease collisions.
These hubs can actually be configured to support either Ethernet or Token Ring. But if they are on the same network they need to be configured to use the same frame format.
10BaseFL/100BaseFL Networks. These networks are based on using fiber optic cabling and the FL stands for fiber link. Fiber optic cabling uses light pulses instead of the electrical signals used in coax and UTP. This results in a much higher quality signal and transmission capacity. It circumvents the resistance that occurs in copper based cabling. Thus, the overall length can be up to 2000 meters which is considerably longer than that for coax and UTP.
Unfortunately, costs are usually much higher and dealing with turns, kinks or any lack of straightness results in extra work and/or equipment.
Token Ring Networks
Token Rings. Token Rings networks were initially developed by IBM in the 1980s. It was developed to overcome some of the problems in contention based transmission protocols like Ethernet. Thus, in general the efficiency or effective use of available band width is typically much better in Token Ring networks relative to Ethernet. But unfortunately, the costs are sufficiently larger to cause most organizations to end up using Ethernet of some sort.
A Token Ring uses a logical ring topology though not necessarily a physical ring. Even though Token Rings are logically a ring they are usually laid out as star topologies. Devices connect to a central hub called an MSAU - Multistation Access Unit. The ring is actually inside the hub, where the wiring connects the ports in a continuous circle and the data travels in a circular path. A signal called a token is passed around the and each client connected to the MSAU cannot transmit until the token gets to it. The token is used both to initiate transmissions and make sure they get to their destination. This way there are never any conflicts on the network. This is how they make considerably more effective use of the available bandwidth versus Ethernets.
The specifications for the Token Ring are defined in the IEEE 802.5 specifications. Usually shielded twisted pair or coax are used for cabling. They are usually more reliable and efficient than Ethernets. If they get overloaded their performance degrades more slowly than Ethernets. They also use an active topology in which devices regenerate the signal to help prevent degradation.
Older versions were 4 Mbps. Newer implementations operate at 16 Mbps. But they are also generally more expensive. You can put a switch at the hub of a star shaped Ethernet network for much less money and get a much higher bandwidth so Token Rings are being see less and less.
AppleTalk Networks. Apple Computers developed a protocol to network its Macintosh computers called AppleTalk. AppleShare is a suite of protocols that operate at the application layers of the OSI model.
AppleTalk networks are often referred to as LocalTalk Networks because the LLAP - LocalTalk Link Access Protocol is the basic protocol used. It also supports dynamic addressing.
These networks can be divided into zones and serve a purpose similar to workgroups in other networks. Users see only resources that are in their zone. Routers use the ZIP - Zone Information Protocol.