The Networking FAQ Q1

Not open for further replies.


[H]F Junkie
Apr 5, 2000
Q. What networking devices are typically used in making a network work the way they do?

A. The devices used in modern networking are as follows:

Hubs are a simple and cheap way of connecting devices such as computers, or hosts, together. Hubs work at the Physical layer of the OSI model and are a simple way to connect machines together. A standard hub, sometimes referred to as a "dumb repeater" allows all connected devices to share its internal single bus, broadcasting all transmissions to all ports on the hub—allowing for a cost-effective solution, but this also means that all connected computers are contending with each other for bandwidth. When connected to a hub, all devices connected to it share the available bandwidth that the hub is capable of providing. For example, if 10 users connected to a 100Mbit capable hub, each user would get 10Mbit each (if everyone were fully using their connection). Realistically, it would be even less since there is a large overhead incurred by checking to see if hub is already in use by another computer before transmitting, or even recovering from when two computers transmit at once - this is referred to as a collision.

Full-duplex (transmitting and receiving at the same time) cannot be obtained using any form of hub. Only half-duplex (transmitting and receiving, but not simultaneously) is achievable using a hub.

There are generally three varieties of hub: passive, active, and switching hubs. Passive hubs are as previous described. Active hubs build upon this, by amplifying and regenerating the signal, allowing for greater cable distances, but at a cost - increased collisions. There is a limit of four hubs per network in every network, due to the collisions expected.

Note: Many times hubs have been referred to as multi-port repeaters.

Switching hubs on the other hand, sometimes referred to as "intelligent hubs", or more commonly as "switches", are wholly different. They carry more onboard circuitry, which gives the switch an added logic; the switch can determine the recipient's port for each package sent along the network, and only the intended recipient receives the transmission. This allows each connection to use its full quota of bandwidth: in the case of a 100Mbit switch shared amongst 2 or more nodes, each node will have a full 100Mbit of bandwidth to another node. Switches are more expensive than traditional hubs, but, naturally, switches
offer far higher performance than hubs, and were traditionally used to connect multiple hubs together; now, they are rapidly replacing standard hubs altogether, also allowing for full-duplex network operation.
In addition to the above differences between hubs and switches, its also notable that switches traditionally operate at the Data Link layer (layer 2) of the OSI model. Recently, capabilities such as routing (layer 3) are now being incorporated into switches, blurring the line between the traditional switch and a router. These “layer 3 switches” are utilized in the manner/location of a traditional switch with added capabilities; hence they are usually referred to as “switches” rather than “routers”. Generally a switch will forward network broadcasts while a router will typically block broadcasts. Generally a switch forwards packets based on MAC address, a router based on IP address. A “layer 3 switch” will do all of these things depending on programming. Whether this is a switch or a router can be debated, but it is used in a traditional switch environment.

There is a catch with switching that needs to be realized. All switches have finite throughput capabilities defined by the bandwidth of their switching fabric. If a switch only has a 1Gbit switching fabric, but has 24 100Mbit ports, it will max out long before offering all it’s connected hosts their full 100Mbit pipe. What happens if the switch runs out of bandwidth? Packets are buffered for transmission, or if the buffers get full, the ports are placed in what is known as a blocking state, which effectively turns off the port until the switch can process traffic again. It is sort of like sitting at a yield sign waiting to get onto the highway. Until bandwidth is available for the switch to transmit for the waiting host, any traffic being blocked or buffered must wait for the traffic to slow down long enough for the packets to get onto the road. A switches’ bandwidth/switching fabric/buffer size determines the cost of the switch. This may help explain why identical port count switches can have costs that vary by thousands (even tens of thousands) of dollars.

Note: Switches have also been referred to as multi-port bridges.

Routers in their simplest form connect networks together. They use the Network layer (layer 3) of the OSI model and don’t really care about anything else but IP addresses. Since routers are used to join two networks together, it should be expected that they use a "higher level" protocol than standard Ethernet networking; such as IP (Internet Protocol). To be true, a proper router is any device, hardware or software, which accepts and forwards traffic inward and outward from a network, effectively joining the two networks. It accomplishes this by learning "routes" to different places on all (adjacent) interconnected networks, thus creating a routing table. When you retrieve a webpage, it is going across many routers to reach the website. At each stage, the router consults its routing table to determine the best way of reaching the site, sometimes based on a single criteria such as the clearest route, fastest route, most reliable route, etc. Routers give the Internet an inherent ability to "heal" should networks fail - they can simply route around missing segments.

Note: Often times a router may be referred to as a Gateway.

For more information on IP addresses visit at

A network interface card, or NIC, is the device in a computer that connects to the system to the network. Modern NICs have a variety of physical interfaces, but the most common one is the RJ-45 jack. RJ-45 is the type of connector that is found on the end of UTP cabling (UTP is further explained in the Cabling section of this FAQ). In a nutshell, its what you would connect into the NIC and Switch to make a computer part of the network.

Bridges (see “switches”) work by joining two local area network segments together. In doing so, the bridge monitors all passing traffic, and forwards it to the other side. Eventually, the bridge learns which MAC addresses are on one side, and which addresses are on the other, creating its own table, and so reduces un-necessary packet transmissions.

Bridges are generally uncommon for use by home users, since the introduction of cheaper, less-featured switches. System and network administrators will sometimes divide a larger network into smaller segments, and join the smaller elements with bridges, in order to reduce traffic congestion known as "swamping". This practice is rapidly dying out, due to the increased use of routers and switches. It should be noted that bridges do not control broadcasts only addressed packets.

If you would like more information on the OSI model, please visit,, or
Not open for further replies.