Welcome to the Markware networking 101 course. Class, please take a seat in front of your computers.
Hopefully you have an understanding of the OSI model that will make this a lot easier to understand.
Hubs operate at the physical layer, layer 1, of the OSI model. You have two types of hubs, active and passive. Passive hubs are hubs that allow computers to connect to the hub and communicate with each other, but do not regenerate the signal and do not have a power supply. You will most likely not encounter these hubs anymore. Active hubs are what you will run into today, but even now some hub manufacturers are moving away from the hub to the more intelligent switch.
Hubs broadcast the incoming frame to all the ports on the hub including the sender. Eventually the frame will arrive at the intended destination and all other computers will drop the frame since it’s not meant for them. This however can bring a fairly large network to a crawl if you are only using hubs and not switches. I'd say use hubs in a networking environment of only 5 or so computers. Since hubs create so much unnecessary traffic, these redundant frames eat up all the bandwidth and create the likelihood of collisions. Collisions occur when two frames basically smash into each other and garble the signal. The collision detection part of CSMA/CD detects a collision and tells the sending computers to backoff for a random amount of time and then resend the transmission (usually a matter of milliseconds).
Remember that this is Ethernet and it uses baseband signaling (i.e. 100BaseTX) so there is only one channel, that is the entire bandwidth of the media being used. Maybe in a later post if you want me to get into WAN technologies, I'll go into broadband signaling.
Ok, on to switches. Switches like their "ancestors" the hub, look very much a like. If you were to take the labels off a switch and a hub and put them up to one another, you probably would not know the difference. However, generally speaking, switches operate at the data link layer (more specifically the MAC sub layer) of the OSI model. There are switches that operate at the network layer, layer 3 of the OSI model that forward frames to their destination by using the logically assigned address, however we assume that we’re are talking about layer 2 switches that forward frames to their destination by using the physically assigned address (MAC address). Once a frame comes into a switch, it checks its MAC address table and sees if the destination MAC address is associated with a certain port (physical port on the switch). If it is, the switch will forward that frame onto the designated port associated with that address and also take note of the senders MAC to see if it is located in its MAC address table. If it is it will say, “Ok! I’ve seen you before.” If not, then it will store that senders MAC and associate it with the port that the frame came in on. Now what happens when a switches receives a frame that it has no clue where the destination is? The switch will now broadcast out for that MAC address, except it will not broadcast out to the sender’s port, and once it finds a port on which that MAC is located, it will then store that information and forward the frame on.
There are three switching methods used by switches. We have the cut-through method, the store-and-forward method, and the fragment-free method. You are all probably familiar with the store-and-forward method, but I’ll explain the others if your not familiar with them. The cut-through method is not commonly used today even though it’s somewhat of a faster method than store-and-forward. Once a frame comes into a switch operating in cut-through mode, the switch reads the header on the frame and immediately begins forwarding that frame on to its destination without any error checking. Now this might seem like a great method, a fast one anyway, except this could begin to be a problem because this means the switch can begin propagating errors. So most people prefer the store-and-forward method since it takes the entire packet in and then forwards it on to the destination. Store-and-forward once was considered to be too slow since the switch had a lot to do on its part including some simple error checking. Nowadays most switches have enough power that it really doesn’t make a difference anymore. Now the fragment-free method uses the fast advantages of the cut-through method, but also checks for fragments. I and probably every other Net Admin on the face of planet earth still prefers the store-and-forward method since it not only improves with troubleshooting, but also is a more reliable switching method. Switches also can operate in full-duplex mode. This means that CSMA/CD is not used and has the potential to double the rate of transmission speed. Switches also create multiple collision domains which means that if a collision did occur, that it would be isolated to only one segment of the network. The process that switches perform is called microsegmentation. There are also managed and unmanaged switches. Managed switches contain an interface with which you can monitor, control, and diagnose the segments on which the switch is connected to or you can use unmanaged switches which don’t have an interface. I use managed switches only when I have servers connected to them and unmanaged switches when I have primarily workstations connected to them. Managed switches are generally more expensive than unmanaged switches.
Let’s move to bridges. Bridges once were commonly used in the days when the bus topology was in its prime. With bus networks however, you have to follow the 5-4-3 rule which is you can only have 5 segments, 4 bridges, and 3 segments can only be populated. However, I will not delve off into the bus topology since no one uses it anymore. Bridges also operate on the data link (MAC sub layer) of the OSI model. Bridges read the header off the incoming frame from say Network A and decide whether or not that destination address is located on Network A or say Network B. If it’s located on Network A, it drops the frame and does not pass it onto Network B or if it thinks it’s on Network B, it forwards it on. It also keeps a table to remember this information if it again receives a frame from that sender or receiver. Bridges segment the network to isolate high traffic segments from low traffic segments. Bridges however have fallen out of favor for switches since most switches already have bridging capabilities. Bridges also have another helpful rule that goes along with them called the 80/20 rule. That means you should only place a bridge where 80% of the traffic is going to stay within that segment and the other 20% of the traffic would be passed on to another segment. Where you physically place the bridges is another matter. There is however a “catastrophic” problem that comes into play when you start using multiple bridges for even greater performance which is called bridging loops. A bridging loop occurs when two or more bridges get confused on where a device is located on the network. Simply put one bridge says, “It’s over there on that segment”, but the bridge on that segment says, “What are you talking about??? It’s over there on that segment”, then that bridge receives it again and says the same thing and the frame gets tossed around getting nowhere. So this “new” protocol was introduced for bridges called the Spanning Tree Protocol (STP) which would tack on a value to the frame identifying which bridge this frame came from. Starting with the
primary bridge with the highest value, other bridges would have a lower value than that of the
primary bridge. If the
primary bridge went offline for some reason, the next bridge with the highest value would be the new primary.
There are 3 different types of bridges. The first one to talk about is the transparent bridge. Transparent bridges are probably the most commonly used bridges out there since the other devices on the network have no clue to their existence. The next one is called the translational bridge. Like its name suggests, a translational bridge can translate the data it receives. These bridges can connect 2 different types of networks together; say a Token Ring network and an Ethernet network. Depending on the direction of travel, a translation bridge can add or remove information from the frames it receives. The final one is the source-route bridge. This type of bridge was developed by IBM for use on its Token Ring networks. The source-route bridge gets its name from the fact that the route of the frame is embedded within the frame. This means that the bridge can determine how the frame should be forwarded throughout the network. With the diminishing popularity of Token Ring networks, you probably will not see this type of bridge in person.
With bridges falling out of favor to the performance and functionally of switches, you probably will be working more with switches than bridges.
The next device to talk about is the gateway. The term gateway is applied to any device, piece of software, or system that can perform the function of translating data from one format to another. The key with gateways is that it does not change the data itself, just the format. Gateways for example can connect two networks together using different protocols. Maybe Network A is using TCP/IP and Network B is using IPX/SPX, a gateway can connect the two and allow them to communicate with each other. Maybe you want to connect an Ethernet network with a Token Ring network; you use a translational bridge which is technically a gateway. Even a dialup modem is considered a gateway since it converts that digital data from the computer to analog signals that can be transmitted over POTS and back to the digital format on the receiving system.
Routers operate at the network layer, layer 3 of the OSI model. This means that they route frames using their logically assigned address. On a TCP/IP network, this would be the IP address. If this was an IPX/SPX network, this would the Internal Network Number (INN). Routers literally route frames around a network. They examine the incoming frame by its header and look at the destination. They then check their routing tables to determine the best path the frame should travel to reach its destination. They take into consideration the link state of other routes, the amount of hops (metric value), and the cost of taking a certain path. Routers make internetworking possible and without them, there would be no such thing as what we call the Internet.
Routers can be a very big topic to cover and if you would like me to go into routing protocols just let me know since this weekend I have a nice mini-vacation
.
thanks so much for going into all that detail. You've clarified most of what I was wondering about, but I'm still a little confused. And btw I don't know anything about the OSI model; it's just very basic stuff I need to know eg. when you'd use what, and the differences.