Hubs  

 A Hub is a generic term that is used for a device that acts as a central point for LAN cable. The most basic and popular types of hubs are devices that simply connect cables together and regenerate data thereby passing data from one device to another. These types of hubs are referred to as concentrators or repeaters. There are also hubs that provide additional functionality, including bridges which connect network segments and routers which connect different types of both local and wide area networks. The functionality of these types of hubs are highly specialized and closely related to the type of network they are used in. They are typically installed in very large, complex networks. As such they are beyond the scope of this document.  

Software  

In addition to the networking hardware previously discussed, software is required by the network to operate. This type of networking software is referred to as a Network Operating System or NOS. The various brands of NOSs divide into two types: those which provide for a Peer to Peer architecture and those that provide for a Client/Server architecture. Additionally, a small software program is required for the adapters installed in LAN nodes to operate with the NOS. These programs are referred to a device drivers or simply drivers. Drivers for popular adapters are often included with the NOS itself. Custom drivers which provide for better adapter performance or drivers for less popular adapters are included with the adapter itself and must be copied into the NOS when the adapter is installed.  

• Network Operating Systems and Required Drivers 
• Developer Product Architecture Type Driver Required 
• Artisoft LANtastic Peer to Peer NDIS 2 
• Banyan VINES Client/Server NDIS 2 
• DEC Pathworks Client Server NDIS 2 
• Microsoft Windows NT Client/Server NDIS 3 
• Microsoft Windows 95 Peer to Peer NDIS 3 
• Microsoft Windows Peer to Peer NDIS 2 or NDIS 3 for Workgroups 
• Novell Netware Client/Server ODI or IPX 
• SCO SCO UNIX Peer to Peer LLI 
• Various Internet N/A TCP/IP 

Network Architecture   
A peer to peer network is an architecture in which all network nodes have essentially equal status. All workstations can directly access data stored in all other workstations. A user on a system configured to a peer to peer LAN would view the data stored on a peer system's hard drive as an additional hard drive accessible to that user's system. For LANs with a relatively small number of nodes, this is a fairly efficient and easily managed network architecture. As the number of workstations grows, it is easy to see how this architecture would become unwieldy. Imagine having 1,000 hard drives to choose from!   

A client/server architecture is an architecture that can be scaled up to a nearly unlimited number of users. With this architecture, the personal computers attached to the LAN are designated as either clients or servers. From a hardware point of view, a server is nothing more than a workstation that is dedicated to serving the data needs of the actual workstations which are designated as clients. Any data required by the client that is not resident on the   

client's system is accessed from the server. If a client wishes to send data to another client (such as an e-mail), this data is routed through the server which then forwards it to the appropriate workstation. The server also acts as a central repository of data files which are accessible to all the workstations on the LAN. For this reason, the server is often referred to as a file server.  

Given the high data storage and transfer requirements of the server, it is typically a very well enhanced system relative to the clients. It will typically have a very large hard drive or multiple hard drives for the network data that it stores. It will have a relatively large amount of system memory and will have very fast input and output devices. As client/server LANs grow, more servers can be dedicated to specific functions like e-mail or a particular database.  

Types of Networks  
Given that the concept Local Area Networking has been established for well over twenty years and given that LANs serve the computing requirements of a wide range of organizations, it follows that different types of LANs have been standardized at various times to suit specific computing needs. The organizations that is responsible for establishing these standards is the Institute of Electrical and Electronic Engineers or the IEEE. This professional organization works with network product vendors and users to standardize product specifications so that hardware and software products from many different and competing vendors will work together on the LAN. Products that conform to these standardized specifications and therefore work together are considered to be Interoperable. Within the IEEE, various committees are set up to act as the standards setting bodies for different types of LANs. These LAN committees are designated by a number.  

The two predominant types of standardized networking are Ethernet and Token Ring. Of the two, Ethernet networks are far more common than Token Ring networks. In general, Ethernet networks are designed for small and medium sized groups while Token Ring networks are best suited to larger groups. Amongst larger networks, it is not uncommon to have both types of networks which are connected via a router or similar hardware device. Given the focus of this document on smaller networks, Ethernet will be covered in more detail than Token Ring.  

Ethernet was developed in the mid 1970's at the Xerox Palo Alto Research Center (PARC). This facility did pioneering research in what are today very popular elements of personal computing like Graphical User Interfaces and mice. Ethernet was standardized by the IEEE by the 802.3 committee.  

Ethernet is predominately used as a baseband transmission network, where all the network nodes share access to the network media on an equal basis. Baseband transmission means that data sent over the media uses the entire bandwidth of the media as opposed to broadband transmission where data takes only a segment of the media by dividing the media into electronic channels. Each Ethernet node has the capability to send data at the Ethernet standard speed of ten million bits (megabits) per second (10Mbps). Since the nodes are sharing the media, the actual data speed tends to be significantly less than 10Mbps in much the same way that the speed of a car on a crowded freeway tends to be significantly less than the posted speed limit. Ethernet is based on a network access method called Carrier Sense Multiple Access with Collision Detection or CSMA/CD. On an Ethernet network, all nodes share the network much like an old party line telephone system. When nodes on an Ethernet network want to send data to another node, the node first senses (or listens to) the network to ascertain if there are other contending nodes already sending information. Any node can send information at any time provided that it first checks to see if the network is already in use. Once the data is sent, the node listens to the network to see if a collision occurred with data from another node that sent data at some time after the sending node last listened. If a collision occurs, the node simply re-sends the data until it successfully reaches its destination address. When there are a relatively small number of nodes on the network, collisions are fairly rare. As the number of nodes increases (as the number of cars on the freeway increase as in our previous example), the likelihood of collisions increases proportionately causing increased re-sends and generally decreasing network performance.  

When Ethernet was first standardized by the IEEE, the type of media specified was thick coaxial cable which allowed for a maximum cable length of 500 meters. Later, a second cable implementation of Ethernet was standardized based on thin coaxial cable that was less expensive and allowed segments of just under 200 meters. To differentiate between these two different cable implementations, the designations 10BASE5 and 10BASE2 were developed for thick and thin coaxial cable respectively. The "10" refers to the 10Mbps data transfer rate, "Base" refers to the fact that Ethernet is a baseband media access method and the "5" or the "2" refers to the maximum cable length between nodes.   

With both 10BASE5 and 10BASE2, the nodes are arranged in a bus topology meaning that they are connected one after another with a terminator on the last node on the bus. While this is a fairly simple design, it does have a drawback in that if a node fails on the bus, it effectively terminates the bus making media access impossible for nodes further down. This is similar to older Christmas tree lights where if one light burned out, the entire strand failed. Trouble shooting a bus type topology to find the problem node is just as time consuming and frustrating as trouble shooting old Christmas tree lights. To address this issue as well as provide for even less expensive media, a third implementation was developed using twisted pair cable and allowing for nodes up to 100 meters from the hub. This implementation was designated as 10BASE-T. Unlike 10BASE5 and 10BASE2, the 10BASE-T physical network design is a star topology with each node connecting not to the next node but to a central hub. If a node fails, the hub partitions that node from the rest of the LAN leaving the remaining nodes unaffected.   

As it stands today, the popularity of Ethernet cable implementations are in order: 10BASE-T, 10BASE2 and 10BASE5. These implementations are not mutually exclusive meaning that an Ethernet LAN can, as an example have a 10BASE5 segment from the file server to the hub, a 10BASE2 segment of older nodes and a newer segment using the 10BASE-T topology.  

• IEEE Implementation Interface (Connector) Type Cable Type 
• 10BASE-T RJ-45 Twisted Pair 
• 10BASE2 BNC Thin Coaxial 
• 10BASE5 AUI Thick Coaxial 

Token Ring is a network architecture that is based on a far more structured media access method. Not surprisingly, Token Ring was developed by IBM for larger LANs. The Token Ring IEEE designation is 802.5. With Token Ring, the network nodes are arranged in a ring pattern. LAN data, along with an electronic "token", circle around the ring. Unlike Ethernet, a Token Ring node cannot send data at any time: it must first capture the token which is constantly circling the ring. The node then attaches the data to be sent to the single token and sends it to its destination. In this manner, only one node can send at a time, eliminating the possibility of collisions. Historically, Token Ring hardware has been more expensive than Ethernet hardware and available from fewer vendors. Today Token Ring is found mostly in large corporate LANs.  

In addition to Ethernet and Token Ring there are other network types with smaller installed bases. Arcnet is an established type of standardized networking that has declined dramatically in popularity relative to Ethernet over the years.   

While Ethernet and Token Ring promise to be the predominant types of networking for the rest of the decade, new, higher performance networking technologies are on the horizon. Fast Ethernet is a 100Mbps version of Ethernet based on CSMA/CD and twisted pair cable. 100VGAnyLAN (another 100Mbps network type) is a sort of hybrid of Ethernet and Token Ring and uses twisted pair cable as well but with a "demand protocol" which structures media access. Asynchronous Transfer Mode or ATM is a high performance network type (25 and 155Mbps data transfer) that can be used for both local and wide area networking. ATM is generally held to be the network of the future, showing a lot of promise but very little implementation at present.