 We've talked about the way to deliver data through a network switching and that we need to find a path through a network Now we're going to look at define Some different types of networks that are used in practice Some of them use switching on a large scale that is a large network Some are much smaller networks which may not require many switches maybe one or even zero so We'll see that these actually are the types of networks. We talk about our Subnetworks in a larger network often that is the internet is built from joining many small networks together What are those small networks lands and and wide area networks? So let's talk about those smaller networks or component networks and And Mainly today it's going to be starting with some terminology Some way that we classify different networks and then some very basic concepts that are common to many networks So some of the issues that we deal with first. Let's try and categorize networks. There are many different types of computer networks in use when we want to compare them and talk about them well, we can classify them in different ways and Some ways we can classify a listed here That is we can classify networks based upon the transmission medium. Is it a wired network or a wireless network? We'll list some examples of each as we go through Will we yes, we'll talk about the trade-offs of of the the difference between a wired network and a wireless network We can classify based upon the link configuration. Does the network use point-to-point links or Point-to-multipoint point-to-multipoint remember one person transmits multiple receive Point-to-point one transmits one receives so the link configuration may be different and we we may have Different designs depending upon whether we use point-to-point or point-to-multipoint and different performance We can classify based upon whether the user of the network the person using the computers to transfer data is fixed or mobile So we can talk about fixed networks The network is not fixed But the users inside that network are fixed the devices are generally in the one position or mobile networks Where the users may move around and therefore we need to design the network to support that mobility The types of users so the users may be you and I the the human users who who are the source of data or the ultimate destination of the data But in some cases we'll have networks that don't support the Individual end users but carry the data of a group of end users So we distinguish sometimes to talk about they're the users where the the end users access a larger network those small networks will refer to as access networks and then the data from those Users accessing the network may be sent via another network and That other network has the goal of just forwarding data of many users a Core network or a backbone network and we'll show a picture to to compare them And the last one is based upon coverage area. How big is the network that the physical size of the network? We'll classify based upon coverage area. So let's look at them briefly What transmission medium should you use for your network? Well? Wired or wireless? What are the advantages and disadvantages of each and you're doing your assignment and you're Doing some experiments with a wired network versus a wireless network and you may see in terms of performance One in general is better than the other What's the difference? Well the pluses and minuses the plus are the positive and then the minus than the negatives and They're almost the inverse of each other that is with wired networks when we have Separate wires connecting our stations together our computers together the interference is very low that is I have a wire going from my laptop to the PC a LAN cable for example and From your laptop to another PC via a LAN cable when we both transmit at the same time Along those separate cables That's transmissions don't interfere with each other My transmission is contained within that cable your transmission is contained within your cable Then there's very small interference the way the cables are designed that the signal stays within there. It doesn't Come out so much As a result with little interference we can transmit signals and What's received is is? Very little noise very little interference So that the good quality of the signal received which translates in the end to higher data rates can be achieved We can get high data rates high speeds Whereas with wireless The second the first negative point under wireless. There's generally interference Not in all cases, but in many cases with wireless networks when my laptop transmits to the access point on the wall and Your mobile phone at the exact same time tries to transmit to the same access point on the wall Because with wireless our signal propagates in all directions normally sometimes we can have highly directional antennas but in in many cases the signals propagate in all directions and the result is that the access point hears or receives the transmissions by from both my laptop and Your mobile phone and those signals overlap at the receiver and The receiver cannot interpret what data I sent or what data your mobile phone sent because it receives a signal which is a combination of both of the transmitted signals and It cannot make out what the data was sent So with wireless because our signals Propagate in all directions Or in not necessary with an omnidirectional antenna, but in multiple directions then They may interfere with other transmissions and interference means that the receiver cannot Understand what was sent which means that less data can be successfully delivered. We get in the end lower data rates The same if if everyone starts talking in this lecture at the same time Then you will not be able to Clearly understand what I'm saying because the transmissions from the other people will interfere with what I'm saying So that's a big problem with wireless With wired we have very little interference. So we can get a higher data rate with wireless. We have interference. So We get poorer performance when we compare them With wired it's easier to upgrade the capacity to get even higher data rates by generally adding a new wire For example, my LAN cable between laptop and PC supports one gigabit per second. I plug it in All right. So the capacity of my network is one gigabit per second. What if I want two gigabits per second? I've gone by another LAN card for both computers and plug a second cable in and Now I've got two gigabits per second between those two devices very easy to upgrade by just making sure we have We we can support a second cable there Because the transmissions across those two cables do not interfere with each other so With wired networks we can upgrade the capacity by just adding another wire or a cable We can't do that with wireless because with wireless we When everyone transmits using the same range of frequencies they essentially interfere with each other How do we add more capacity in that case? Well, the only way to do it would be to use different frequencies That is I can transmit to the access point at a rate of 100 megabits per second with a wireless technology Then if there's a second person who wants to transmit to that same access point at 100 megabits per second because of interference Effectively, we'll only get half of that 100 megabits per second each I will get to transmit 50% of the time the other person will transmit 50% of the time getting 50 megabits per second each That's the problem with interference. We effectively divide that capacity by the number of users How do we upgrade? We can't just add another access point because we still use the same range of frequencies What we can do is use a different range of frequencies a different channel. I Transmit to the access point using channel one you transmit to a different access point using channel six if we think of Wi-Fi and The signals were then not interfere because we're using different range of frequencies That's a way to increase the capacity but We're using more Bandwidth in total to do that more bandwidth incurs more cost We can't just add an infinite number of frequencies. We're limited on the range of frequencies. We can use so With wired we can get higher data rates and it's easy to upgrade to increase the speed add a new wire and Related to no interference that the delay is usually small across cables and usually predictable Every packet I send has about the same delay and you may see that when you run your experiments across your Land link when you ping the other computer that the values reported are about the same all the time With wireless because there is interference. I may send a packet It may may not get there. I have to send again retransmissions incur extra delay But what most many wireless systems including Wi-Fi design to do is that because there is interference They follow a protocol such that only one station transmits at a time Again my laptop wants to send to the access point your phone wants to send to the access point Then they follow a protocol such that the aim is that if we both want to send only one of them will transmit at a time Maybe my access point transmits first and then your phone transmits If there are 10 devices, maybe they can take in turns and The problem with that is that for my device's perspective I get to transmit and then I have to wait for everyone else to transmit. So there's a delay For when I can send my next packet at the end today We may talk about if briefly the ways that we can take in turns allow other device to transmit so with wireless The poor performance may include varying delay and Again, I think you'll see that if you do a ping across your wireless LAN those values reported in ping Will get different Delays we can even try that now I'm connected via Wi-Fi from my laptop to the access point and I'm pinging from mine to another Computer on the third floor 10 10 6.1 We see the delays are in the order of milliseconds But some of them jump up here. We had about two The average is around six, but we have one here 15 milliseconds Several at one millisecond. This one was 69 milliseconds three milliseconds. So we see the variance there that's because When my laptop wants to send the message then The protocol that the wireless technology is using is designed such that of all the devices in the area that want to send Only one will transmit at a time. So sometimes mine will have to wait for your phone to transmit So I want to send But there's some delay and waiting for others to send. So that's why we may get The higher delays maybe that at that particular time other devices were wanting to send Sometimes I get to send first so I get a small delay of one millisecond So that will depend upon how many other devices are wanting to send if we did it across a wired LAN You'll see that it's about constant through through those different pings What else about wired versus wireless? What's the problem with wired? It's expensive to install especially when in locations where it's hard to access Okay, that's we have to install the cables So if we've got a building and we want to have network coverage Then it may be difficult to put the cabling through the building rather than we don't want to lay it on the floor So people chip over it or it looks bad We want to put it in hidden locations and that can be difficult in some cases Or we want to cover across the city We want to have wired length links across the city then we have to get Permission to dig holes under the under the roads under the the land owned by different people and that's an expensive operation The other problem with wired is that the devices are typically fixed. I plug a LAN cable into my laptop I can't move it very far. Maybe within a couple of meters, but but depending on the length of the LAN cable If it's wireless, I can move my laptop Essentially anywhere that provides coverage from the the wireless access points So wired leads to fixed networks normally Wireless allows for mobile networks mobile users inside the network And that's the key advantage of wireless no physical connection. We allow mobility and it's convenient Even my laptop just sits here. It's not really mobile I don't have to plug in a LAN cable to get the network access. It's convenient another problem with wireless systems is that Because many people want to transmit using the same frequencies we need to have some rules about who can use those frequencies So often there's a license required to use those frequencies So with mobile phone systems, you can't go and build your own mobile phone network put up some base stations You must get a license to use the range of frequencies that the mobile phones will transmit and receive on So the mobile phone operators pay a lot of money to get those licenses With Wi-Fi unlicensed That is we don't have to pay a license to use Wi-Fi So there's some range of frequencies which are classified as unlicensed But with those the more people using it the lower the performance I will get So that's the problem with unlicensed bands that anyone can use it the more people the lower the individual performance Last thing wireless physical security is difficult With a wired network inside SIT the LAN cable going from the desktop on here to the switch downstairs for someone to listen in on the transmissions So we transmit a signal from the desktop to a device downstairs Along the cable for someone to intercept that data They need physical access to the cable what they could do if they want to intercept the data on my wired network They need to come into the building and maybe put a special tap or intercept on the the LAN cable And then they can intercept everything I send ignoring encryption So to get access to a wired network from a security perspective an attacker needs physical access to the network They need to come to the cable to the building and so on but with wireless When my laptop transmits up to the access point the signal is also going outside So maybe someone standing outside the building not even on our campus May receive the signal transmitted by my access point By my laptop and also by the access point in the opposite direction So there's no physical security there That is an attacker doesn't have to be inside the building to intercept the data that I send So it's much harder to provide that physical security and wireless systems So we need to make sure we use encryption So that even though someone can receive the signal they cannot understand the data that was transmitted And we'll cover that next semester in the security course So that's our first choice. We've got wired and wireless networks and we would choose the medium depending on our requirements Any questions on wired versus wireless? Especially since your assignment is really comparing wired and wireless So you you're studying that a lot. Let's get through the others a bit faster The link configuration point-to-point That is a pair of users have their own link a dedicated link And as a result, there's only two users using that link so that the performance They'll get is is usually high and predictable. There's no one else that we need to share that link with Therefore, we know what performance will get you when we transmit across that link When we have point-to-multipoint then We usually require some way for sharing that medium for sharing that link so that we Allow each user to transmit and wireless is a good example there Again Everyone wants to transmit their data from their laptop their mobile phone up to this access point You all want to transmit data using the same frequency at the same time If you all transmit at the same time All of your signals will interfere and the access point won't understand anything that any of you send and that's bad So we have what's called medium access control protocols such that Your phone and laptop follow some rules with the aim that only one of you will transmit at the same time So the rules will be right this device gets to send Then the next device gets to send and then another one gets to send and so on so that they don't create interference That lead lower performance Because you have to wait for all the others to transmit you'll get lower throughput higher delay So that's the problem with point-to-multipoint links that because multiple people may transmit and receive We must share the medium amongst them. We don't get dedicated performance for ourselves And there's some complexity in doing this medium access control which leads to some overhead What's the problem with point-to-point if we want to cover many users we need many links There's some cost involved and we need to plan the endpoints So if we want to connect everyone in this room Then we'd need many cables to connect everyone together so that we can all talk to everyone else With a point-to-point point-to-multipoint system Maybe if I want to communicate to everyone else then I transmit and with a wireless system everyone can receive That's very easy point-to-point doesn't always Relate connect with wired links. So wired links are commonly point-to-point But we can have wireless links when we use directional antennas. They are effectively point-to-point on the top of the other building on our campus we have an antenna pointing to our rungset campus and It's a highly directional antenna which essentially creates a point-to-point wireless link Those devices when they transmit only ones receive because the antennas align such that the signal will just go in that direction We will got not get interference from others Point-to-multipoint wireless links like Wi-Fi mobile phones But also there are some wired technologies that use point-to-multipoint communications bus type technologies old lands used to be such that My desktop will transmit onto the LAN cable and everyone would receive that transmission So in the old LAN Old ethernet it was a point-to-multipoint topology We'll see some pictures and examples of different topologies shortly the other classification was with respect to users and here's a simple View of a large network which is made up of many smaller networks So think of we have one large network The the PCs of the users the sources and destinations and any PC wants to communicate with any other Well one common setup with large networks is that we use Some technology for the user to access that large network and that Network itself we refer to as an access network That's where the end user gets access to the larger network But and there may be multiple access networks Then we have some other networks To connect those access networks together So there's different naming for those Intermediate networks here. We may call it a core network So these two access network connect via this core network and we may have multiple core networks And maybe a backbone network that connects all of them together The point here is that the end users Access the larger network via the access networks The core and backbone networks don't have end users directly attached. There's no Person with sitting in their computer Directly connected to the core network the core and backbone networks are designed just to transfer the data From access network to access network The result is that the designs of those networks the technology selected may be different The core networks and backbone have to be designed to carry a large amount of traffic a large amount of data So we need high-speed links. We need multiplexing Often reliability mechanisms are important The access networks and need to be designed to make it easy for the individual users to connect an Example may be Wi-Fi is an access network or a wired LAN or a mobile phone system or ADSL and then the networks operated by the the SIT by Internet service providers telecom companies they are the core network and backbone networks using wired LAN optical fiber Maybe dedicated point-to-point wireless links. So that's another classification of networks last one How big is the network and here we often hear some acronyms that Are a measure of how big the networks are how much area do they cover and there are many different names here and there's no Defined standard as to the names how big they are, but I've given some rough Measures of the area that the networks cover and some example technologies So we can start some networks are built to cover a small area in the order of centimeters and Maybe it's a network to connect devices on your body that is connect your mobile phone to your earpiece Using Bluetooth for example or connect multiple devices that you carry with you not necessary on your body, but nearby Or maybe it's just for connecting on your desk to connect your laptop to your monitor wirelessly an example of a small area network a Body area or personal area Using technologies like Bluetooth, you know, I think you've all probably used Bluetooth or you've got your devices that support it But there are others in for red technologies Zigbee is a technology for similar to Bluetooth, but for low rate low data rate communications for things like Reporting status data, so not having to send a lot of data But over a small distance and you can use a wireless connection, which doesn't consume much battery power So the battery can last not for a day like your mobile phone, but can last for weeks and months and There are other wireless technologies Primarily wireless technologies that support covering across very small area Body area networks personal area networks at some of the names Then if we move up and think about how do we connect across? inside our home inside an office or an office building Then we're talking about networks that cover meters several meters tens of meters. Maybe even hundreds of meters in some cases Well often they refer to as a local area network the network is covering just a local area Land, okay, so that's the main Acquire them that will see a rise Sometimes if it's specific to the home it may be a home area network or maybe if it's for specific for storing data That is a company needs to store data On a number of different storage devices then those devices may be connected via a network storage area network What technologies are used there? Ethernet our wired land technology The formal standard created by the organization called IEEE and the number of the standard the document that describes Ethernet is 802.3 This organization IEEE has Created standards for many different wired and wireless technologies in the land space and also wide area networks So they created the standard for Ethernet IEEE 802.3, but also for Wi-Fi IEEE 802.11 is the standard by the same organization that says how do I transmit from my laptop to the access point wirelessly and within these standards there are many Variations and improvements if you've in your assignment looked at your Wi-Fi access point You'll see that it may support different substandards 802.11a.11b, G, N, AC They are just substandards or variations of the original and they're primarily designed to support across a local area Sometimes they used to build larger networks, but the main design and motivation to support In homes offices and buildings And there are some others as well They're not the only two Then we move up and think about how do we build a network that covers across a city that connects between cities across countries and eventually between countries and Across cities referred to as metropolitan area networks More generally across cities between countries a wide area networks a WAN lands and WANs talking about distance of kilometers of kilometers hundreds of kilometers, maybe even thousand and these networks are often Core networks and backbone networks So wide area networks are often run by one company and that that company Really leases or rents the access to that network to others that have access networks Because they are maybe very expensive to run and operate then they those wide area networks are built To carry that the data from different access networks So the access networks maybe via lands Which then connect to a wide area network to transport the data across a larger distance Lowes networks need to be designed to carry a lot of data Not just the data from one or a few users, but the data from tens of or hundreds of users or thousands of users So they need to be high capacity They need to transmit that data across long distances and They need to be quite reliable Imagine the network that transports the data from One of the internet service providers in Thailand like True True corporation has a as a gateway that connects to other countries. So there are links from the true International internet gateway in Thailand that goes to Singapore to Japan Malaysia and other countries those links are essentially wide area network links or wide area networks and They need to carry the data from not just you and I but from thousands of users If the link fails or the network fails then thousands of customers are unhappy So reliability is an important part of these large networks They may have backup links or backup paths so that they can automatically switch over to a backup link So that even if one part fails the rest of the network can deliver the data So there are different technologies used there. Many of them are wired technologies, but there's also wireless technologies Some of them listed here PDH and SDH are Wide wired technologies really for for high capacity links across cities between countries If you look on the later slides in this Lecture, which we won't get time to go through there's some listing of the the data rates for PDH and SDH so they go from tens of megabits per second up to 10 10 or more gigabits per second and They are commonly used say for links between countries And and across a country and there are some other technologies atm It's not about money, but it's about a way to transfer data using packet switching virtual circuit packet switching and frame relay was an older version and then some wireless technologies Ymax is a Some there's some relationship with IEEE 80211 Wi-Fi it's allows usually long-distance wireless links often point-to-point and Satellite is another option Satellite access net a satellite networks That's covers between countries. What if we want to cover the whole Earth Or even go beyond Earth Then we're talking about what megameters 10 hundreds thousands of kilometers Even more Well, we can think that may be a global area network the internet Which is really that made up of connecting many different lands and wands together That's what the internet is connection of multiple different networks covers the globe and if you go beyond the earth the connections to the To satellites the connection to to spaceships The around Mars and other other spaceships we can think of an interplanetary networks We will not look at those So we are going to focus Primarily on lands which cover the the home office building type size networks and Mention a little bit about wide area networks, which are connecting across cities between cities between countries any questions about these Classifications of networks Just be aware that we talk about Networks from different perspectives medium link configuration the user coverage area Did I miss one? We didn't talk about user mobility. Okay, that is some network support users being mobile Some may assume that the users don't move Another comparison of wide area networks versus local area networks and now when I say a wide area network I Will I also mean a metropolitan area network which is covering a city and when I say a land I also mean a home area network or a storage area area network which are really subcases of a land and in fact Sometimes a personal area network is close to a land as well. We may have similarities so Comparing it from a different perspective of WAN versus a land Some of the differences Wide area networks connect devices or networks over a large geographical area Wide area for example between campuses the links between bunkety and rung sit we can think the technologies a Wide area network Between office buildings. Okay, if you have offices in different locations and between cities and countries Often those networks are owned and operated by organizations on behalf of the users That is the networks are owned and operated by internet service providers or telecom companies those owners of the network are Not the end users of the network. What they do is they build the network own and operated and Rent or lease out access to other companies and organizations For example, SIT. We have a wired link or a path from between our two campuses and We we pay another company to operate that link. So another company owns that link They they built it. They they maintain it. We just pay them to carry our data across their their link So they own the wide area network, but we it's our data that carries Across that link that link may carry other people's data at the same time So the wide area network may be designed to not just carry one users data, but the data of many users So they at least access to those networks at least to other users Universities companies may be smaller internet service providers Local area networks usually connect the end user devices Within a campus buildings homes Often not always, but often the land is owned and operated by the organization using the network What I mean is that the wired land inside our campus is owned and operated by SIT and The users of that wired land are the the SIT desktops and computers Okay, that is the end users are part of the organization that own and operate the network and that has a difference on in terms of What options we have to select the technology? That is if we're going to build a new network We are the we're the person using that network. So we choose the technology best for us With a wide area network, we have to rely on the other organizations to choose the technology We can just go to them and say I want to access you from this location to this other location What do you have on offer? We don't get to choose the the network technology Typically, not always again lands will support higher data rates than wide area networks Now that's a bit strange when we look at some examples. What I mean is well one example you are aware of Anyone have ADSL internet access whereas used it at home or somewhere what data rates do you get with ADSL? Anyone know Eight megabits per second. So in the order of several megabits per second, maybe one Down upload maybe eight or ten up download you can get faster So think of that to our part of our wide area network connection Because in fact for us to access the ADSL We use usually Wi-Fi or wired land. You may have a the router which gives you access inside your home How fast is your Wi-Fi or wired land? How many megabits per second? You're doing the assignment on it. How fast you is your wired land? 100 megabits per second Wi-Fi ranges maybe 54, but that's maybe quite old now most new devices support More than a hundred megabits per second with Wi-Fi There's an example that our link to the through the ISP Is only eight or ten megabits per second, but in certain internal for the land We can get a hundred hundred plus megabits per second The land gives higher data rates than the the WAN the wide area network why? Well traditionally if you think of now SIT the land supports one gigabit per second But our link from SIT out to the internet is not one gigabit per second You may notice if you start to download a lot from outside of SIT Then it's maybe several megabits per second Tens of megabits per second Internally we have a 1,000 megabits per second the connection out is in the order of tens of megabits per second So again much higher internal than the external link. Why? traditionally many Organizations a lot of the data was internal communications Okay, inside of business. I'm not always accessing public internet Websites and so on often inside the business. We're accessing internal servers. We're communicating internally Therefore we need enough capacity to support internal communications plus external communications the link Outside only needs enough capacity to support the external communications not internal so traditionally the the land had a higher data rate and higher needs than the external link Nowadays it's changing a bit in that a lot of applications even for businesses are not hosted internally But are hosted out on the internet and therefore a lot of communications is via the external link now even though SIT's link from our campus out to the internet maybe in the order of 10 megabits per second the Cable that carries SISD SIT's traffic out may support a much higher data rate Because it will maybe carrying the data of multiple users. We may use multiplexing So maybe there's an optical fiber network owned by to T that SIT uses and We send our data from our campus to rung SIT But that network also carries data from other organizations So the network may be the WAN may have a Lower per user data rate but higher In terms of the total network capacity So that's just a rough comparison between wide area and local area networks let's look at For both of those types of networks some common topologies The topology is how we arrange the nodes and links Let's say we want to build a network. We have many devices many end user devices We want to allow each end user device to communicate with every other end user device That's a common goal that if I have a hundred desktops in my In my network, I want to allow anyone to communicate with anyone else So how do we connect those devices together? What links do we create? Well, that's the topology and there are some different choices Now we'll talk about the the devices that the end users use that create data and The destinations of data sometimes we call them simply stations hosts or n nodes But in some networks, we have some intermediate devices that will support the data transfer and They may be referred to as switches repeaters and hubs Especially when we talk about lands those names will come up. So let's look at Four or five different topologies that are common that we can choose from and we'll look at it from the perspective of lands But in fact they also apply for wide area networks and note that the links We'll be able to choose from as either point-to-point or point-to-multipoint So we'll go through mesh bus ring and star and a hybrid is simply combining To or more of the above together Before we look at them. What's our requirements? What do we want? We should allow any station to be able to communicate with any other station if we don't that's not a good network design Generally using point-to-point links gives us better performance than point-to-multipoint links So when possible use point-to-point links, but there may be some limitations of them So if I have a point-to-point link and the capacity is a hundred megabits per second Then we can transfer at that full 100 megabit per second between the two devices if it's a point-to-multipoint link with a capacity of 100 megabits per second then the performance each user gets depends upon the number of users We'd like to use as few links as possible think of the links may be cables The less we use the easier it is to install the easier to upgrade the less cost involved We'd like a network that scales well. What that means is when we add new nodes new stations That it's easy to add them For example, I've got currently 100 nodes in my network. I build the network. I want to add another 10 more It should be easy to add those 10 more. We shouldn't have to change the current network to add 10 new ones We'd like the network to be fault tolerant That is I've got my network and there's a failure one link fails not all links, but one of the links fails Then maybe the two devices connective either link cannot communicate But the other devices still should be able to communicate if one link fails if one link fails and That causes no one to be able to communicate then maybe that's not a good network design similar fault detection Sometimes it's nice to be able to automatically detect if there is a failure That is a link fails How do I know that link failed? Well, firstly that the users of that link may come to me and say your links not working. Please fix it Maybe that takes a minute or hours before I know that automatic detection would be the link fails and Immediately there's a packet sent to some special server saying this link has failed and then maybe some correction Action can take place and get a new link fault detection is Very important for large networks wide area networks the link fails Automatically the different link is used as a backup So given those requirements, let's compare four different topologies that are common first one's mesh and the example is we have six stations a BCDEF and We want to connect them using links and a mesh topology We use point-to-point links and have a link between every pair of devices So we see the mesh here a full mesh in that Station a has a point-to-point link to be to see D E and F and Similar B has links to every other device This is good for performance with point-to-point links everywhere Then the performance between each pair of nodes is guaranteed That is assuming every link is a data rate of a hundred megabits per second A can be sending data to be at a hundred megabits per second and at the same time to see and at the same time to D E and F Those links are dedicated for the the two Endpoints There's no sharing. That's good. What's the problem with a mesh topology? cost why Why is it costly? Many many cables many links. All right, you see in the picture. There are many links necessary. How many links here? Come on count them for me. How many cables do we need? Did you count them? How many lines between the devices? Sorry a has links to five others B has links to four others C has links to three others To D to two others and E to one other that is we don't count them twice five plus four plus three plus two plus one is 15 There are 15 links in this case Let's grow it to a real-size network say inside SIT. We want to connect. We have a hundred devices We want to use a full mesh. How many links do we need? Well the general formulae that the number of pairs of devices how many are there? It's n times n minus one over two That is here. We have six devices the number of pairs is Six times five divided by two thirty divided by two fifteen With a hundred devices the number of pairs the number of links is a hundred times ninety nine divided by two Which is about five thousand? Okay, we need it less than five thousand So I think you realize the more devices we add The more links we need and that becomes very costly to deploy and actually difficult to run all those links through the the space Of course the other problem every device needs multiple links plugged into it The device a has five cables plugged into it So we need a network interface card that supports those five links So that's not very practical good for performance not good for the practical purposes of deploying those links and Making the devices simple So not commonly used in large networks only suitable very small networks Maybe when we need very high performance like in small wide area networks We want to connect devices and we want good performance between every pair. Maybe a mesh is okay a full match Note that we don't need any form of addressing when we use point to point links when I send something on the link Who does it go to? Well, it goes to the other endpoint I don't need to say in the message that I send who is the destination because there's only one possible destination We'll see in the other approaches. We will need some form of addressing So we can identify who is the destination So when I transmit a frame from A to B That frame contains some data. We don't necessarily need to include the address of the destination in That frame because the destination if a transmits it always will be B on this link There's only one possible destination So mesh is possible, but not very practical for for large networks. We could have a partial mesh Not connect all of them together That saves on our links. What's the problem? Can a talk to F? well Assuming that the other devices do not forward data then a cannot talk to F Assuming the role of the devices is either to create data to send or to receive data for themselves I will not send data for other on behalf of other people Then in this case a cannot talk to F and we set our requirement. We want to allow everyone to talk to everyone else So that's the problem with a partial mesh topology that we may not allow some pairs to communicate If we do allow some nodes to connect to forward data That is if I allow for example D to forward data What I can do maybe every node of forward data to send from A to F I could send from A to B B realizes I'm not the destination So sends to D which then sends on to F. That's our switching We've covered that in switching. We can do that if B and D are switches In terms of packet or circuit switches we can deliver the data from A to F But if they are not switches, they're just end stations like a laptop or PC We can't communicate between NF This may be used in some wide area networks when we don't necessarily need to communicate between all pairs Let's look at some other Topologies primarily used in local area networks First one is a bus topology What we do is we have a special link That the six devices attached to and that special link refer to as a bus and You've studied bus topologies or you may have heard of them in computer architecture when you think we have Between the components inside a computer they can communicate via a bus communications bus the same concept here what we do is that when a wants to transmit it has an attachment to this bus link so this solid line is the Cable you think in practice is a cable that passes every computer and Each computer attaches has a special attachment to that that one cable So there's one long link and everyone has very even though it's shown to be long for D But has usually small Devices that attach to that cable and What happens when a wants to transmit to F a transmits onto the bus and The bus usually is designed so that the transmission goes to everyone attached to it and Everyone includes F so F Everyone receives a copy of that transmission and We include the address of the destination inside the frame So when a transmits it sends a frame destination F The bus delivers that frame to everyone connected to the bus B gets it notes. The destination is F and ignores it See gets it and ignores the frame if gets the frame and accepts the frame because it is the destination This bus link is a multi-point point a multi-point link One transmits multiple points receive It's not a point-to-point link. This was used in early a wired lands Okay, so before we had The current land topologies what we'd have is a long cable that passes by every computer usually desktops in those days And the desktops at the back of the land card will have a special attachment that connects on to that bus link and Your computer would transmit everyone would get a copy But only the one who was the destination would take a copy the others would discard the copy So that's why we need addresses in the frame The the bus link had some special terminating points such a To end the link it was rather the advantage is it was rather easy to install We don't need multiple cables like in the mesh topology. We just need one Link passing every device So let's say we have an office building what we would do is we'd have a long cable that just runs past every computer And each computer would attach on to that connect to the bus So not hard to install Now the link the bus link is a point-to-multi-point link the problem with point-to-multi-point links is that We must share that medium amongst multiple users. It's not dedicated for just a pair of users the result was that Only one user could transmit at a time if a is transmitting data to F Then B cannot be transmitting data to E at the same time That bus link is shared so only one user would transmit at a time The result is that if the link capacity is a hundred megabits per second On average the amount that each user can get is that 100 divided by the number of users 100 divided by 6 in this case 16 megabits per second so with a point-to-multi-point link Roughly we can say we divide the capacity by the number of users using that link If we had a hundred devices attached and the capacity was 100 megabits per second roughly each user would get one megabit per second each We don't get the full 100 each and So to share the medium required protocols to do that automatically and it resulted in lower performance One problem was if the bus fails no one can communicate and That was so an issue if there was some error here that normally then it stops everyone from communicating and that's not good fault tolerance There were some practical limits on how long the bus could be and how many devices could attach so in the old lands There were some limits. You couldn't just have one long bus that covers Thousands of meters through a building so this was used in earlier lands, but not very common anymore It was replaced by the star topology, which we'll see Similar to a bus almost the same, but we connect the two endpoints So instead of having that that one link with terminating points connect the two endpoints of the bus and you get a ring topology same concept a transmits onto the ring and The frame is transmitted all the way around Where each device gets a copy of that transmission and the one that is indicated as the destination Takes a copy and processes it the others ignore the message So very similar to the bus topology Note that usually the transmission would happen in one direction Essentially a transmits and it goes that the signal goes past f f gets a copy and the signal keeps going past e E gets a copy and so on comes back to a and then the signal Terminates or that the frame is removed from the ring Again similar to the bus not hard to install compared to a mesh especially it was One of the advantages is it was easy to identify faults That is if there's some portion of the link that's not working Then by sending a message around we'd identify which Device or which portion of the link is not working. That is if this the message always goes Clockwise we go from a and it goes around to f and back to a If we send a message and the link from D to e that portion is not working or D is Or e is not working if we send a message around it will get to D It will not get to e So we've now identified which portion of the link has has a fault D got it e didn't get it so that must be there in the link and that's a Important feature in large networks where we want quick notification of faults Especially in wide area networks covering a city or a country We have a ring network and if there's a fault we can quickly detect that and fix it It was used in very old lands But again not very common except where those old legacy networks are still in use The concept of a ring topology is used in some larger metropolitan and wide area networks This fault tolerance or fault detection is an important feature Larger networks would have a ring in each direction And even backup rings I transmit Around this direction if I want to send from a to f I send in a ring in the opposite direction because it's closer And if there's a fault if we detect the fault from D to e we can switch over to the other Other link the other ring so we can quickly recover last one And especially for lands very important today because what we use in wide lands is a star topology We introduce a new device in the network We now have our six stations a through to f And we introduce a new device which is not a station but as an intermediate device It gets different names But the idea is we have point-to-point links between station and this intermediate device And we want when we want to communicate From a to f we transmit across the point point link to the intermediate device Which then sends across the point point link to f This is like switching Remember in our packet switching we'd send to an intermediate switch Which would then send to the next switch well here. We just have one switch And a common name for this intermediate device is a switch and in a land a land switch or an ethernet switch The benefit in this case is that we're using point-to-point links The ring and the bus topology. We have multi point links and we got lower performance per user here So long as the switch is good quality that is as fast enough I can be transmitting from a to f At my full data rate And at the same time be transmitting say from e to b At the full data rate If my links support 100 megabits per second I can be sending at the same time 100 megabits per second from a to f And 100 megabits per second from e to b And 100 from c to d Whereas with the shared links that wasn't possible So much higher performance That does depend on whether the this switch is fast enough to to process that traffic And nowadays they usually are The problem with this is that we've introduced a central node if this node fails Our whole network fails. No one can communicate So we have a dependence on the central node, which is a problem If the link from a to the central node fails Of course a cannot communicate but everyone else can still communicate So that's okay fault tolerance For that with respect to the links If one link fails it doesn't prevent others from communicating That wasn't the case with the the bus and the ring If the bus fails generally no one could communicate So that's a good thing with respect to fault tolerance Generally easy to install We have one central device and from every station we have a one cable going in there So with a hundred stations we need a hundred cables With our mesh with a hundred stations we needed five thousand cables Usually the links actually were full duplex Or or two links that is What we could do if a is transmitting data to f At 100 megabits per second At the same time with full duplex links f can be transmitting data Back to a or to someone else At 100 megabits per second If we have full duplex links then we can do that and that's what we have in most lands today So this is used in in the lands that we use today This pc has a land cable plugged into it It goes into the the socket in the wall and that land cable goes down to a switch the central node Down in a cabinet downstairs And every other pc connects into that same central node and we get this star topology There are Over time there are variations on what this central node did And the first versions it was a what was called a hub And The hub was designed such that really only one device could transmit at a time The hub was very simple But then as the the hardware got better and cheaper People created a true switch which allowed the device To be able to support the data from a to f and at the same time from b to e and at the same time d to c So this central device in today's lands is called a switch An ethernet switch or a land switch You may have heard of hubs but hubs are usually No longer used because it's the same price to buy a switch as a hub And a switch is faster So that covers the main topologies we'll see in lands And some historical And also we'll see them in in wide area networks as well And that's the main thing we want to cover in this topic We've introduced and classified different types of networks by different criteria and listed it some of the the common topologies that are for connecting Devices together The last two things in the slides at the end. There's some example network technologies You can look through and see the names of some The the topic on medium access control we will not Uh, I think have time to cover This is about the ways in which we share point to multi point links When we have multiple users wanting to send at the same time We said that if they all send at the same time, we will interfere with each other So we want to avoid that And the ways to avoid that Is Well three basic approaches Fixed assignment involves either giving the users different frequency FDMA means frequency division multiple access We've talked about FDM and TDM Same concepts, but with multiple users If there are two users wanting to transmit at the same time give them different frequencies And with different frequencies they can transmit without interfering or TDMA allow them to transmit at different times First user transmits then second then first then second. So that's TDMA time division multiple access SD S DMA space division multiple access put them in different locations So that when they transmit they don't cause interference Interference happens at the receiver when two transmissions have are received at the same time With space division multiplexing say with wireless systems you can have antennas So there's two transmitters want to send to me at the same time If I have two antennas one pointed at the first one And the other antenna pointed at the second user they can both transmit at the same time With my two different antennas I can receive the two signals without them interfering That requires some directional antennas CDMA code division multiple access requires some extra processing to allow Some special coding to allow them to transmit at the same time The other one which is of interest because it's used in Wi-Fi is random access Don't control who transmits just let them transmit randomly So Multiple users want to transmit to the access point at the same time Or they follow some rules that says when I have data to send I will check if anyone else is sending I'll listen in is anyone else sending now No one sending then I will transmit And everyone follows that with a little modification that When I want data to send I will check if someone else is sending and I'll wait a random amount of time And I'll wait if no one's sending then I will transmit And everyone does that And if it's designed well, it allows them to transmit One at a time because If two users wait a random amount of time, they'll usually wait a different amount of time One will get to transmit first and when that one's transmitting the other user will Detect that the first one's sending Random access is used in Wi-Fi and was used in some of the old Wildlands but not in the current wildland But especially important for Wi-Fi because it's a very simple way to allow multiple phones laptops computers To transmit without interfering with each other The key thing for you to remember without knowing how they work with random access especially Because we only transmit one user at a time Then we must share that medium amongst those users If my data rate is 100 megabits per second I have 10 users sharing that medium On average each user will get 10 megabits per second 100 divided by the 10 users And in your performance experiments with the assignment I hope to maybe some of you will see that performance If you try multiple users But that's we're not going to cover how that works What we'll do is next topic you'll look at Wildlands and ethernet a few of the details of ethernet We will stop there. We will not go any more through this topic. The next one is on ethernet