Excerpts from an extensive technical reference:
LAN and WAN Subnetworks Under IP - Lan Interconnection
by Thomas A. Maufer

You may hear techies talking about their telecommunications hardware. These devices are critical to letting a human being use a Web-enabled client to quickly and seamlessly communicate with another human being somewhere else on the planet. And, more and more so, they are enabling groups of people to come together serendipitously and do fun and creative things on their communications devices (of which computer, cell phones, and tablets are just some flavors). The question of the week to ponder is: which hardware device is good for what? Let's take a look. The goal is for you to have a basic appreciation of these devices and then perhaps be willing to join in on the conversation next time a friend's spouse brings up one of these devices at a social event.

We'll start with the most basic hardware device: the Hub.

Hub

A hub is a signal repeater – a device that is the simplest device to implement within the networking OSI model layer. A hub is a piece of hardware (comprised of lots of small switches and signal paths – typically copper wire but not at all limited to that medium) that enables a common connection point for many devices in a network and are commonly used to connect segments of a Local Area Network (often called a LAN as if it were a common word in the English language). A hub receives incoming data packets from devices that are connected to it via a telecommunication link (a data packet is just a efficient packaging of a piece of a message – a message being anything that needs to be communicated like a piece of spoken text, a video picture, written text, or instructions to another piece of hardware that lets more than one person play a game coherently – we will be looking at data packets closely next week). The hub's only requirement is to forward those data packets onto any other telecommunication links that are connected to it physically. It therefore repeats the message on to others – most importantly doing it reliably and without altering the message (not to say altering the message isn’t a goal of some communications – think a translation service for example – but that is outside of a hub’s scope of responsibility (it simply acts a data link at level 2 of the OSI model)).

If you hadn’t seen a hub before, you’d be quite disappointed when you did. A hub is just a slab of plastic with multiple data link connections and some blinking lights. The key here is that it does an important job within Web architecture faithfully – we can never underestimate how important that is (think about the player who holds the ball for a field goal kicker – critical to the success of the goal of kicking a ball between two poles). The hub is just one part of what can become a more and more elaborate system that can support valuable processes to quality of life on this planet. The beauty of an architecture – allowing for specialization of mass produced and simple devices that can be designed together into elaborate systems.

Before we move on, just make a note to yourself that a basic hub does not perform any filtering or redirection of data. It does support a most basic synchronous (at the same time) chat room when everyone who is chatting has a telecommunications path to the hub – where everything that everyone types in the chat room is seen by everyone else (a valuable service in its own right if the group uses it to discuss something of value to the group or others). If there are too many people trying to chat beyond the capabilities of the hub, things get bogged down and messages aren’t received in a timely (and acceptable) manner. You pay more for a more powerful hub – a more powerful hub just means more telecommunications connections that can be reliably and faithfully serviced on a timely basis.

A passive hub serves simply as a conduit for the data, enabling it to go from one device (or segment) to another. So-called intelligent hubs include additional features that enable a system administrator to monitor the messaging traffic passing through the hub and to configure each port in the hub based on that traffic level. Intelligent hubs are also called manageable hubs. A third type of hub, called a switching hub, actually reads the destination address of each packet and then forwards the packet onto specific outgoing telecommunication links.

In beginning networking newsgroups, you frequently see a question like "Why can't I just plug all the computers and the cable modem into a hub?"

Connecting the cable modem to the uplink (the one telecommunications connection that goes out to your internet provider’s network) or any other connection on your hub won't work unless your Internet Service Provider can provide multiple Internet Protocol addresses to you. (This will make sense in its entirety by the end of our class together if it does not yet now.)

The reason that you need multiple addresses is that the hub is a repeater, not a router. To review, in simple terms, a hub just takes the data that comes into a port and sends it out all the other ports in the hub – it doesn't perform any filtering or redirection of data among different networks.

Since you often need to create a separate network to isolate one LAN from the rest of the Internet, you need to filter the traffic between the networks and not simply pass everything on your local network to the greater Internet network. The hub doesn't provide the filtering necessary to keep the networks apart from each other – it repeats everything to the outside world whether you wanted to share it or not.

Extra techie talk here (for you to research on your own by searching the Internet): There are special messages that software can produce that connects the uplink port on the hub to a cable or DSL modem (typical ways home users connect to the outside world) in a way to create two networks. This port is just a convenience provided on some hubs so that you don't have to use a special kind of cable (called a crossover cable) to connect multiple hubs together or connect some devices to the hub. The uplink doesn't perform any filtering or routing.  The uplink port is just a specially wired port that has the connections for the receive and transmit channels reversed. This allows connecting hubs together (sometimes called daisy-chaining) to provide more ports, or to connect a hub to some kinds of network devices (like cable or DSL modems).

Bridge

A bridge (sometimes called a transparent bridge) also works at OSI model Layer 2. This means bridges don't know anything about specific messaging protocols (agreed upon standards on how to communicate that we will look at next week), but just forward data depending on the destination address in each packet (again, the data packet is just a convenient way to organize part of a message). The destination address is not the Internet Protocol address (again, we’ll know a lot about that soon), but the MAC (Media Access Control) address that is unique to each network adapter card. Remember that one of the most important responsibilities of OSI layer 2 is to provide a unique address of each device on the network – an IP address is a software address that can refer to one or many devices (so, we need physical addressing schemes for each piece of hardware to use to identify itself as a unique device – an MAC address is the most popular way of doing that).

So, the bridge is the device that is used to connect two local-area networks (LANs), or two segments of the same LAN that speak the same networking language (meaning, share the same protocol) – like a physical bridge in a highway network that connects to street grids on different sides of a river.

With a bridge, all your computers are in the same network subnet (the term meaning a coordinated group of telecommunications connections that work together as a part of the greater Internet), so you don't have to worry about not being able to communicate between computers or share an Internet connection. However, the only data that is allowed to cross the bridge is data that is being sent to a valid address on the other side of the bridge. If the data packets aren’t addressed to a valid address, those packets are not allowed to across the bridge (think Checkpoint Charlie in Berlin that kept capitalists from crossing into communist territory – ha! I think it usually worked the other way around).  Bridges are implemented completely in hardware (physical ports and inter-connections between physical ports) don't require programming.  They learn the addresses of the computers connected to them by listening to the data flowing through them. A port is the generic term for the interface one physical device makes to another physical device – a phone jack (where you connect a telephone wire to a wall) is one familiar example of a port. The word port is used extensively in networking language – and includes connection opportunities made in software within a server (to be discussed further below).

Bridges are very useful for joining networks made of different media types together into larger networks, and keeping network segments free of data that doesn't belong in a particular segment.

More helpful home networking techie talk: DHCP servers will work fine across Bridges, or if you assign your own IP addresses: you should then use the same first 3 "octets" of the IP address for all devices on each network being bridged (Example: 192.168.0.X)

Switches

Switches are the same thing as bridges, but usually allow multiple connections with the same flavor (A very popular example of a flavor is 10/100BaseT). 

Switches can be used in heavily loaded (the load refers to the amount of traffic on the network) communications networks to isolate data flow and improve performance. In a switch, data between two lightly used computers will be isolated from data intended for a heavily used computer, for example. Or in the opposite case, there exist "auto sensing" switches that allow mixing of 10 and 100Mbps connections seamlessly so that the slower 10Mbps transfer won't slow down the faster 100Mbps flow (Mbps refers to the bandwidth or capacity of the communications channel in millions of bits per second – the bps refers to the number of bits per second that can flow down that path – and a bit is one binary value of 0 or 1, on or off).

Although switch prices are dropping so that there is very little difference from hub prices because most home users get very little, if any, advantage from switches, even when sharing broadband (connections with a high bps compared to dial-up, low, bps communication connections) Internet connections.  Broadband connections for most users are in the 1Mbps to 2Mbps range, far below the 10Mbps speeds that the hubs and switches can handle. Since you share that bandwidth within your network, you can see that a speedy 100BaseT connection wouldn’t even breaking a sweat when you're using the Internet on a home network (since it could handle 25 to 50 such broadband connections being used to full capacity – something we haven’t been doing with typical Web applications – but the popularity of video applications is starting to change that).

The switches are the key devices being used by your ISP (or phone company) to manage thousands of connected users to the Internet at once. If you draw a simple map of your neighborhood and imagine everyone connecting to the same phone company and using a video service to watch movies over the Internet, you can get a sense of how many 100Mbps-capable switches would be needed if each 25 home users used up its capacity. But, remember most people also have another cable coming into the home from a cable company – the race has been on as to who can build the better switching empire to service neighborhoods reliably with necessary performance to keep the video (and voice track) running smoothly.

Router

Routers also forward data packets from one place to another too – however, routers are OSI model Layer 3 devices, and forward data depending on the network address from the layer above, not the hardware (MAC) address from the layer below (this is a key distinction that should eventually have huge architecture ramifications to you if not already). For the TCP/IP networks we will discuss in two week’s time (the most popular networking type of the 1990s and first decade since year 2000), this means the router manages sending data packets (and thus communications between users) according to the IP address of the network interface (and not the physical address).

Since we haven’t really discussed much about layers 3 and 4 yet, the following might not make that much sense to you yet, but it has to be written (and you need to come back to this distinction when you have more knowledge later in this course):

Routers isolate each LAN into a separate subnet, so each network adapter's IP address will have a different third "octet" (Example: 192.168.1.1 and 192.168.2.1 are in different subnets).  They are necessary in large networks because the TCP/IP addressing scheme allows only 254 addresses per (Class C) network segment.

Bottom line is that routers, like bridges, provide bandwidth control by keeping data out of subnets where it doesn't belong – they just do it in a more intelligent manner using software and/or more sophisticated hardware. As a result of the needs of the additional intelligent processing, routers need to be set up before they can get going, although once set up, they can communicate with other routers and learn the way to parts of a network that are added after a router is initially configured.

Routers send lots of messages to each other to improve their intelligence as the Internet does its overall job of coordinating messages as best as possible based on varying loads and communication link outages (an outage means an expected service like a communication link is not currently functioning as expected). I find the beautiful design of router messaging to be one of the most enjoyable thoughts of any technical thoughts I have ever been introduced to. So, I hope to share that vision with you as we build up our skills in this class!

At this point (with hubs, bridges, switches, and routers in place for our use in sharing any communication message between addresses in a network), we can start to look at those devices that actually work with us humans to make those messages more meaningful to us humans that produce and consume then. We break the set of all such devices into two main categories: Clients and Servers. Routers are becoming more and more server-capable so that they muddy the characterization of devices. But, there are specific things servers do that classify them as servers.

Servers

A server is a computing device that stores communications content and serves it to requests for that content. If I use the Web to play a video game, I request the software that lets me play the game. The server maintains a directory of all possible software that I might want to access and serves it to me when I request that software. The server can work at any level from level 4 to 7 of the OSI model – which means it needs to connect to connect to the devices we’ve already covered in this document in order to perform the job it wants to accomplish. The devices, written about previous in this document, handle getting a request for video game software from one device on the network to the server (therefore handing the layers below layer 4 on the server’s behalf).

All kinds of content can be served. As you already probably know and enjoy, servers can serve audio, video, text, games, business applications, e-mail messages (and anything else that can be stored digitally as content). The sky is the limit for creative and attractive content to be served to everyone on the planet (note your huge responsibility in this, artists!). Files are stored in traditional file systems that are basically identical to your home computer’s file system – the difference is the file systems are extensive and often span multiple physical storage devices to support a single server. Servers address files the same way your home computer does – so you already know a bit about the structure of file systems: typically very hierarchical (for example: C:/my_name/my_folder/my_file.ext).

Typically, only techies are interested in the services in a server that perform levels 5, 6, and 7 of the OSI layer. There are many fascinating applications that let a network administrator review which files have been served when and to whom (by the IP Address of the requestor). There are fascinating applications that let a network administrator charge fees for access, restrict malicious users, perform better load balancing, clean up unused files, etc. Almost anything you would want to know about a server is made available through server software that runs at OSI layer 7 to make sense to the administrator. You might be surprised how interesting server application data is if you could see it via some of the very pretty pictures that have been made by artists to represent the enormous amount of data being served worldwide (hint: we need artists to help us see and respect that enormous messaging space that impacts our thoughts and beliefs so dramatically these days).

We’ll see much about what a server does at the lower levels in the next three weeks so I’ll stop there for now. Let’s just say that a lot of money has been made by server designers and lots of money is made by competent administrators who keep them running reliably and free of criminal activity (and malicious activity in general).

Clients

OK, artists and Web developers. A wonderful cadre of thinkers, designers, and moneymakers has given everything above this paragraph to you. We need you to join the fray! You can help work on the device that has the most important responsibility of the whole enterprise: making the messages that fly around the network more meaningful and comprehensible: the client.

A client is a hardware device that a users of the network uses to communicate to other users. Too often, the client has been used to communicate asynchronously (not at the same time) with other clients by one client storing information on a server and another client downloading it later. The Web has been set up to let us communicate synchronously (at the same time) but we rarely do it outside of video games, which by their nature, push the technology to keep up with high messaging loads and make the technology more capable for other applications – applications you and I can design and convince venture capitalists to fund!

Useful notes from a past student on the subject:

On searching online about hardware devices, I found this link: http://computernetworkingnotes.com/comptia-n-plus-study-guide/network-devices-hub-switch-router.html

It was really helpful to me looking at the photos of these devices. It made easier to identify them. So, I wanted to share it with everyone!

Here is my interpretation from this week's subject:

I think that having the appropriate devices in place is critical for your computer to work efficiently. Therefore, it is essential to understand the importance and the function of each one of them, and choose them wisely.

HUB:

• It is a repeater, not a router. It repeats the message (without altering) on to others in a reliable manner.
• Does not perform any filtering or redirection of data among different networks.
• More powerful hub means more telecommunications connections that can be reliably and faithfully serviced on a timely basis.
• There are 3 types of hub: passive (serves a conduit for the data); manageable or intelligent (has additional features); and switching (reads the destination address of each packet)

BRIDGE

• Also called Transparent Bridge.
• Forwards data depending on the destination address in each packet.
• Destination address is not the Protocol address, but the Media Access Control (MAC) that is unique to each network adapter card.
• Used to connect two local-area networks (LANs), or two segments of the same LAN that speaks the same networking language (shared the same protocol).
• Only data allowed to cross the bridge is data that is being sent to a valid address on the other side of the bridge.
• They are implemented in hardware — physical ports and inter-connections between physical ports.
• They learn the addresses of the computers connected to them by listening to the date flowing through them.
• Useful for joining networks made of different media types together into larger networks, and keeping segments free of date that doesn't belong in a particular segment.

SWITCHES

• Similar to Bridges, but usually allow multiple connections with the same flavor.
• Can be used in heavily loaded communications networks to isolate data flow and improve performance.
• Key devices used by phone companies to manage thousand of connected users to the Internet at once.

ROUTER

• They are OSI Model Layer 3 devices.
• Also forward data packets from one place to another. Depending on the network address from the layer above, not the hardware (MAC) address from the layer below unlike the devices previously mentioned.
• Like Bridges, it provides bandwidth control by keeping data out of subnets where it doesn't belong in a more intelligent and sophisticated manner.
• Needs to be configured in advance.
• They are becoming more server-capable

SERVERS

• Stores communications content and serve it to requests for that content. It maintains a directory of all possible software that I might want to access and serves it to the user when requested.
• It can work with any level from 4 to 7 of the OSI model

CLIENTS

• It is a hardware or software that accesses a service made by a server. It has been used to communicate asynchronously with other clients by one client storing information on a server and another client downloading it later.
• Per Wikipedia: piece of computer hardware or software that accesses a service made available by a server. The server is often (but not always) on another computer system, in which case the client accesses the service by way of a network.

Address Resolution Protocol (ARP) for hardware device identification

The principle of ARP is extreme simple. For example, the entire interaction process of ARP only requires two packages: one query and one response! However, the ARP protocol also has its confusing places for beginners, such as "proxy ARP", "gratuitous ARP", and "reverse ARP" extended from itself, and these different types of ARP are applied to different scenes. Before diving into the technical principles, as a beginner, let's remember the following three points:

1. ARP is used to implement mapping from a networking IP address to a piece of hardware's Media Access Address (MAC) address, to query the MAC address corresponding to the target IP address.

2. In network communications, the data packets communicated between devices need to be encapsulated from top to bottom according to the OSI model, and then sent out after the data encapsulation is complete. Therefore, in communications of the local area network, not only the encapsulation of the source and destination IP addresses is required, but also the encapsulation of the source and destination MAC addresses.

3. In general, upper-layer applications care more about IP addresses than MAC addresses because they depend on them. Therefore, it is necessary to obtain the MAC address of the destination host through the ARP protocol, and then complete the data encapsulation.