 So in looking at the internet we arrived last week at introducing the concept of an autonomous system where we said it's a network that is under the control of one organisation. So in the internet, the internet is made up of multiple different smaller networks all connected together via routers. Some of those smaller networks are owned and operated by one company, one organisation. The collection of those small networks owned by one organisation we refer to as an autonomous system. Most of those organisations are internet service providers. But not all, we'll see Thomas Hart University has its own autonomous system because we have our own set of networks that connects to the internet. Before we go back into autonomous system let's look at a very simple example of internet and internet routing. You do not have this one, so I'll put the two or three pictures here, I'll put them on the website but for now just try and follow along. It's not so much, it's a bit of a review for some of you, that is some of you should know this already. Here's a very simplistic example network from say an internet service provider in Thailand. A small ISP that has coverage across part of Thailand for example. They have a network in Bangkok and maybe they have a network although it's not shown in Chiang Mai and other parts of the country as well which would look similar to what I show here for the detail of Bangkok. Of course a real ISP may not have a network structured like this, it may be much more complex but just to give a simple example, up the top I've drawn four different networks or sub-nets and let's say they are the four access networks. You as a customer, you want internet access, you go to this company and you subscribe and you connect via one of these networks, A, B, C and D, so if I drew a computer, your home computer it would connect maybe via an ADSL link into this network A, so the cloud refers to some sub-net, some network and I've given them labels so we can refer to them, so network A, B and so on and then because they are spread across Bangkok in different parts of the city, those networks or let's call them sub-nets to be clear, these sub-nets A, B, C and D are connected together and they are connected via a set of routers, so R1 through to R4 are routers. So remember a router connects sub-nets together, so subnet A connects to subnet E via router R1 and E connects to G via R5 and so on. If I'm a customer connected to sub-net A and I want to communicate with someone on sub-net D, what path would my IP data take? I'm on A, I want to communicate to D what path will it take or what path may it take and let's define the path by a set of routers. So what we do in the internet is the host, the one who creates the IP packet, sends to a router which sends to another router and so on until it gets to the destination host. So the path from a host on sub-net A to a host on D is again R1, R5, 6, 4. That's one path in this case, just in this simple scenario. So that's the path that our data must take to get to D because there's no alternatives unless we go out and come back in. In each of the routers they are configured to store information about that path. So when I send a packet, I send it to my first router. I'm on network A, I send to R1. Why do I send to R1? I have a routing table inside my computer that says to send to anyone on the internet, you've only got one router to send to you or default router. Send everything to R1. R1 receives my IP datagrams and then looks in its routing table. And its routing table should be configured to say that if anything is destined to sub-net D, R1 should send to R5. That's how the routing table in R1 would be configured. Let's just summarize or write an example routing table for R1 and keep it simple. If the destination is on sub-net D, the next router I send to, if I'm R1, should be R5. Simple. Consider the destinations of the other sub-nets, B and C. What should the entries be? If I want to send to someone on network B, if I'm at R1, I should send to R2. And to C, R5, and I'll reach eventually. And assume then each of the other routers are configured with a routing table such that when it gets to R5, if the destination is D, goes to R6. If it gets to R6 and the destination is D, that should send to R4. If the destination was C, R6 should send to R3. So the routing tables are configured to store information such that path is taken for our IP datagrams. What about, although I haven't drawn them, there are similar access networks in Chiang Mai and Chamburi and elsewhere. For example, LM and N are in Chiang Mai. Routing table for R1. I want to reach L, subnet L. Where do I send to? If I'm at R1 and I want to reach a subnet out here, R5. Okay? MR5. Okay? So we're drawing the routing table just for R1. Anything out here, to these other cities, in this configuration from R1, we send to R5. Okay? Because there's only one path to reach out in this case. So we could simplify, instead of listing them all here, we could simplify the routing table and say, at least conceptually, we could... I'll just do it for this one. Anything else? Star means any value. If a destination is D, send to R5, B, R2, C, R5, anyone else who is outside of Bangkok, send to R5. So that's a way to simplify the routing tables. Of course, we could simplify it even more because we've got two instances of R5 here. This is a simple case in that there's only one path. What if there were multiple paths? When we need the same thing, we need to select a path, which is best. And the creation of these routing tables in real networks is not done manually. Someone doesn't sit there and calculate them and type them in. The routers run some software that implement a routing protocol. And the routers automatically exchange information saying what networks they are attached to. And from that information, they calculate the paths and populate the routing table with the next hole. So this generation of the routing table is done automatically inside our network. It could be done manually here, but if things change, we need to update it. And things may change quite often in some networks. So paths, if there were multiple paths, then we'd have to deal with that. For example, let's say this new network is attached to a new link to Chiang Mai. It doesn't show the destination, but from R6 to Chiang Mai, as well as this one. So now there are two paths. So for example, before, if we were at C, or routing table for R6, before we had to reach Chiang Mai via R7, now we can reach via this path. So as things change in the network, our routing tables need to be updated. And normally inside a network, the update is automatic, performed by routing protocols. So that's what happens inside an organization's network. And it's up to the organization to choose which paths, or at least to choose the policy or the metrics for selecting the best paths. So that's routing inside an organization's network. Routing protocols calculate the paths for us. This ISP, our example Internet Service Provider, we see R10 here. I call it a border router, although it's not shown. There are links to other networks here. It's a border from this ISP's network to some other ISP's network. So we're just looking at one Internet Service Provider in this diagram. There may be others, or there are others beyond R10. And if we move up to a higher level view, let's replace all of this ISP's network by just a single cloud, ISP1, for example. So instead of drawing all the details, I just draw a single cloud and say, this is the network owned and operated by ISP1. Inside a multiple subnets, A, B, C, and I haven't drawn them here, up to T, for example. Just to keep track, there are multiple subnets inside ISP1's entire network. So I just designate that here. Let's keep track. Inside ISP1's network, there are, in fact, multiple subnets from A through to T. I give them some names or some labels. And then, of course, there are other Internet Service providers or other organizations in general that own their own network. So focus on connecting between different organizations' networks. Here we have our router R10. We call this a border router. It's a border between one organization's network and another organization's network. It connects them together. So in this case, ISP1 connects to ISP2, who, inside here, have their own networks, and which then connects on to ISP3. So we use a different protocol to exchange information between ISPs. And the protocol we use is called BGP, the border gateway protocol. Inside here, inside this, there are multiple different routing protocols we can choose to update the routing tables. But between two ISP networks, usually the protocol used is called BGP. It is different from used or it implements, it works differently from the protocols used inside networks. BGP is used between the organizations, more specifically between autonomous systems. And what they do, in very simple terms, is that using BGP, the ISPs tell each other what networks they have inside their network. That is, ISP1 tells ISP2, inside my autonomous system, I have networks A through to T. And ISP2 tells ISP1, inside my network, I have networks A, a U through to Z, if we give them some label. Similar, ISP3 tells ISP2 about the networks it has, it ran out of letters. So AA, ABA, AC and so on up to AQ. So BGP is used to tell, to exchange between autonomous systems, the set of networks that are inside there. So then, if someone on ISP3's network wants to send to our subnet user on subnet A, so the one someone up here, then internal for ISP3, they have their own routing protocol and they have their own routing tables inside the routers to deliver from the source to the border router, R20, and then that border router knows that to reach subnet A, it needs to send to ISP2 and ISP2 will deliver across their network and because of the exchange with BGP, ISP2 sends on to ISP1, which eventually delivers to the destination host on subnet A. So we've got two levels here, inside an autonomous system, inside one of these clouds and between autonomous systems or between these clouds here. Inside we use one of many possible routing protocols. It's up to the organization to choose. Between them, everyone uses BGP, the border gateway protocol. I'll say a bit more specific about IP addresses, but any questions on this simplified view of the internet and autonomous systems? BGP is the border gateway protocol. It's a protocol that all organizations use to exchange information about what networks are inside their network. So ISP1 wants to tell ISP2, I have networks A, B, C, D up to T, and similar ISP2 tells ISP1 about the networks it has available. So when someone on a host, on network A, creates a packet and the destination is on network, what do we have? Z. Then the routers, the host on A, the destination is on network Z. This host will send to R1. R1 should be configured. The destination is not D, B or C. The destination matches something else. Therefore, send to R5. Similar R5 should send to R7. And internally, the packet should be delivered to the border router because we know that Z is not internal. It's somewhere outside. And the border router knows that subnet Z, Z is on ISP2's network. How do they know? Because these two ISPs exchange information using BGP about what's on their network. So BGP is to exchange information between the autonomous systems, between the ISPs, about what subnets are inside their network. Of course, I've only drawn three ISPs here. In the Internet today, there's something like 40,000 different autonomous systems. And you don't just have a connection to necessarily one. ISP1 may have connections to multiple other ISPs. So when the destination is subnet Z, it must know to send to ISP2, not to some other ISP. And how does it know that? Well, because BGP told this border router to reach Z, send to ISP2. If the router had a link to other ISPs, it wouldn't send in that direction. Any further questions before we move back to our lecture notes? Okay, in your lecture notes, and we've mentioned that, let's call there are two types of protocols. There's interior routing, or instead of routing, gateway is sometimes used. Interior gateway protocols, they are used for routing inside an autonomous system. So there's an IGP, an interior gateway protocol used inside here. Some examples of specific interior gateway protocols, and they're listed on your slides. In fact, we'll not remember many. Some examples of interior gateway protocols are RIP, OSPF, and others. There are many to choose from. The company chooses whichever one suits their needs. But between autonomous systems, we use an exterior gateway protocol. And the one used today is called BGP, the border gateway protocol. So think of BGP as a specific instance of an exterior protocol, RIP, OSPF, and so on, instances of an interior protocol, gateway protocol. Think of IGP and EGP as types of protocols, not specific ones. So Thomas Hart, TU has its own autonomous system. In fact, we'll see an example shortly. Let's say they choose to use OSPF inside TU's network. But then between TU and true internet, we use BGP to exchange information. Although we don't use letters to identify networks. That's just a simplification in my example. What do we use to identify subnets? What do we use? What do we use to identify a subnet in the internet? A subnet mask is part of it. Or we use an IP address and the subnet mask. It's the IP address, a specific form. In the IP address, we have 32 bits. Some of those bits identify the subnet. The rest of the bits are used to identify a computer in that subnet. So the IP address is used to identify subnets here. We don't have network A. We have a network 203.108.16.0 and a subnet mask. That refers to a network instead of the letter A. And when ISP2 exchanges information with ISP3 about what's inside its network, it's a little bit more than that. It's more about what is reachable from my network. There are no other ISPs. Let's say there are more ISPs in this direction. From ISP3's perspective, which subnets are reachable via R20? Well, the subnets reachable are U through to Z, but also A through to T. Because to reach here, we need to go through ISP2. So it's not just exchanging information about what's in your network, but what's in your network and what's reachable via your network. So we could say the subnets reachable that can be reached via ISP2 are U through to Z and A through to T. So now when ISP3 has a packet and the destination is network B, then it should know to send via ISP2. Because they exchange information and ISP2 tells ISP3 networks A through to T and U through to Z are reachable via my network. So we'll send in this direction. And this will send to R10 and eventually to subnet B. So it's more complex than just about what's inside your network. It's more about what's reachable via your network. And since we do not use letters, we use IP addresses. The structure of IP addresses and how we use subnet masks makes that possible or easier. I'll show you an example after we look at TU's network a bit later. Let's return to our slides and look at some aspects of autonomous systems and then we'll look at some examples for real autonomous systems. So in summary, an autonomous system, a group of networks under control by one organization, one administrative authority. Usually an internet service provider, a large company, a university, a government organization. And each autonomous system is assigned an AS number, an autonomous system number. And that's managed by some organizations. The central organization delegates that to regional Asia Pacific Network Information Center, for example. And we mentioned this and we finished last week looking at the Thai internet map. And you can see when you look at the Thai internet map, you can see the autonomous systems for different ISPs, the AS numbers. Inside an autonomous system. So in my previous pictures I referred to an internet service provider, but more generally we call it an autonomous system. So inside an autonomous system where you can use one of many interior gateway protocols, between autonomous systems we use just one exterior gateway protocol called the border gateway protocol, BGP. And the neighbor autonomous systems, they advertise what's reachable via their network. And this is a different example where we have our multiple autonomous systems. One, two, three, four, five, six. Why have I got five, five and five? That's a mistake. It should be five, six and seven. There's not three in the same number. They should be different numbers here. And zoom in on AS three and we just see these different subnets here. So this is inside the autonomous system. We use an interior gateway protocol. Between we use BGP. The remaining parts we want to discuss is really about the relationships between autonomous systems. How they exchange traffic and something about how they pay for that traffic. Because if I'm going to, I, as a customer, as an end user, pay my ISP to carry the traffic I want to send, but then they need to pay someone else to carry their traffic, to carry my traffic. So there's some commercial relationships between the ISPs. We'll talk about some notation and some terminology about that. But before that, let's look at an example. I'll zoom in in a moment. This is just a website by some internet service provider, Hurricane Electric. There are other websites. They provide some information about the autonomous systems, about the routing in the internet. So it's a nice website to give us some information and some examples. I visited the website and from, when I visit the website, it detects who sent that packet, who is the source IP address. And the source IP address is 20313120966. And then it identifies from the information they have in their database that that corresponds to Thomas Art University. The specific network it corresponds to is 203131208.0 slash 23. And my ISP is AS37992. So TU has its own autonomous system. It's not an ISP, but many universities also have an AS. Let's zoom in and then look at some details. So let's look at the autonomous system for TU. I will not try and explain everything here, but some things that we can recognize I'll explain. This is the information about it and scroll across. It talks about prefixes originated and announced, BGP peers, AS paths, and the average path length. The prefixes really refer to the subnets here. So 20310816.0 slash 24 is a prefix that is the network portion of the IP address, the portion that identifies the subnet. So within an autonomous system, you manage certain subnets. In our example, ISP1 managed subnets A through to T. It advertises or announces those networks A through to T to its neighbor autonomous systems. Well, referred to as prefixes. Instead of sending the letters A through to T, they send IP address prefixes. And in this case, TU announces 26 in total prefixes. We'll see a list in a moment. And peers, BGP peers observe. So the peers are the neighbor autonomous systems. From TU's perspective, there are two neighbors. In our example, we saw each ISP1 had one neighbor, but ISP2 had two neighbors. So TU has two neighbors as shown here. Let's look at who those neighbors are. We just focus on IP version 4. The peers for TU's network are true and uni-net. So we're connected directly to two other autonomous systems. All the traffic out to the internet via TU's network goes via one of these two networks. True, as you know, is a commercial ISP. Uni-net is a network that connects multiple universities together, run by the government. In fact, both of these we can think are internet service providers. TU has connection to two. You don't have to connect to just one. You can connect to multiple. Some of your traffic can go via one and some via another. If we try and draw that, it's a TU's network, which is not just SIT but the whole university. So there's many different sub-nets inside here. There's one for our campus, there's one for SIT at Rungside, and probably most of the different faculties have their own sub-net. So inside the TU network. And then we have connections to two different autonomous systems. One is by True and one is Uni-net. And then, we'll see shortly, they have connections to other autonomous systems. And our traffic, so when you visit the YouTube website, for example, when you're here on the TU network, it goes to one of these two networks, let's say via True's network, then goes to another autonomous system and eventually reaches the autonomous system that has the YouTube web service. We scroll across. There's just no need for you to record the numbers, but just for our example, this was 37992, autonomous system number. True is 7470, 4621 is Uni-net. We'll click on them in a moment. Let's see what else we see for TU. Who is, tells us information about who owns or operates a network. So if we scroll down, this is the data about the TU autonomous system. Who assigned the number and specifically here. The name, something about who they connect to. They import and export data between two different ASs, 74770 and 4741, which is different from here. So this, in some cases, this could be out of date. Things change over time. I don't know which one's out of date. 4741 or 4621, one of them may not be current anymore. It depends how often they update their database. Something about the owner of the network, the contact details for TU's network. And so what happens is TU tells its neighbors, True and Uni-net, about what prefixes, what subnets are reachable. And in fact, they tell other neighbors, neighbors of neighbors as well. So they propagate that information to multiple different autonomous systems. And the graph shows us who TU propagates information about their network to. Here we are, TU, our direct two neighbors, we tell, but they also inform their neighbors, or some of their neighbors at least, not necessarily all of them. And that's what this graph shows, that information about TU's network is known by all of these autonomous systems. And that's Singapore Telecom, Singtel, NTT, Savas, Sprint, AO and others. So there are some other internet service providers that information about TU's network has reached these other autonomous systems. That's what this graph is showing. And let's look at these, our two neighbors, and see about our neighbors. So one of them, let's first look at Uni-net. Let's first look at the summary information. So Uni-net, which is an organization that connects the multiple universities together, connects, has 30 peers. So TU had two peers, two direct neighbors. Uni-net, this network has 30 different peers. So it exchanges information directly with 30 other networks, that's what's shown here. And it exchanges information about many different sub-nets, many prefixes, up to 200 different prefixes. And the list of peers of Uni-net are here. So CAT, a Thai research network, and multiple universities listed here. And go back and let's look at true internet again. Version 4 and V4 and V6, is that what you mean? IP version 4, IP version 6. So there are two versions of the internet protocol. The main one you use today is IP version 4, that's what you use most of the time. But the newer version is IP version 6. It's not as commonly used yet, it is available. So some networks use it, some don't. So in these cases, they separate between the information exchange on the IP version 4 network and the IP version 6 network. That's what the V4 and V6 means. We're focusing on V4, IP version 4. The last thing here in this example, let's look at true's network, which is 7470. So similar to true, which is a large ISP has connections to 65 other peers. So it connects to multiple different peers here, 65 here. And they exchange information about their networks with many other autonomous systems. So we can see some of the peers listed here. The primary one being the true international gateway and many of the other ISPs within the country. So from this, and there are similar websites, we can see something about the connectivity between the autonomous systems in the Internet. And you can see which ones are connected to each other. To look at the details, you need to understand more about the classless addressing and especially BGP, which we will not cover in this course. Any questions on autonomous systems? The last thing we want to cover is who pays who. Who pays? Who do you pay for Internet access? You pay one Internet service provider. You pay to use their network. But they don't have a network all across the globe. When you access YouTube's website, say the servers, there may be some in Singapore or in the US, your ISP most likely doesn't have a network that connects directly to the destination. They only have, for example, a network inside Thailand. So your data goes through multiple ISPs' networks. So the arrangement of how you pay your ISP to carry your data, then there needs to be some arrangement between your ISP and the next ISP between the peers here as to who pays for what. So let's look at some of the terminology related to that and how the connections can be made. So two autonomous systems connected together we've referred to as peers. So TU peers with True and TU peers with UniNet. We talk about peering. So they are peers with each other at the same level. To connect between peers, you need a physical connection and some commercial agreement. So the physical connection, some links, some cables, and the agreement is some contract that says, I will pay you this amount of money for the amount of megabytes that you carry of mine. We'll talk about both of them. The physical connection, two types. A private peering connection where two companies get together and say, let's join, let's connect our networks together. For example, True has a router here and TU connects their cable into that router. So some physical connection between two peers directly. The other way for a physical connection is via some public peering point where there's some organization that provides the routers and multiple different peers connect physically into that organization's network. Usually called an internet exchange point, an IXP or simply IX. So one way we draw that and it's similar with the True, the internet gateways in Thailand, multiple autonomous systems. Of course, one option is to directly connect them. So this one has a router. They make an agreement and connect the cable to there. So they peer with each other via direct private peering. But another option is that there's some central point, some building or some data center and they all connect into there. And then from there they can create an agreement with whoever they like. So this is the physical connection. They all connect into some central point, some network here and therefore they can easily send their traffic from AS2 to AS1 or AS2 to any of the other peering points so long as they have an agreement. So in terms of the physical connection, you either directly connect or you connect via some shared point called an internet exchange point. Let's look at how the agreements work. So often it's a commercial contract between the two organizations established to agree upon how much data they're going to carry and what they're going to pay for it. There are two general types of agreements and the terminology we use or as common is transit and peering. But be careful, there's some confusion in the terminology used here. Peers refers to the two autonomous systems that connect together. They appear. The physical connection, we can have either private peering. They connect directly together with each other like here or public peering where they connect via some shared exchange point. That's about the physical connections. But then we talk about the agreements, the contracts. We have a transit agreement and a peering agreement. So don't confuse a peering agreement with public peering. We can have public peering with a transit agreement. We can have public peering with a peering agreement. So distinguish between an agreement and the physical connection. What's the difference between transit and peering? Transit is when one organization pays another for them to carry their traffic. You pay that other network or that other operator to have your traffic go through their network. That's a transit agreement. That's usually the case when, let's say, we have a large internet service provider and a smaller ISP. The smaller ISP pays the larger one to carry their traffic. So I may subscribe to the small ISP. They only have a network in Bangkok just across the city. I subscribe to them. I pay them. But then for their traffic to get out across Thailand and across the world, they pay a larger ISP for network connectivity across the country and across the globe. So we say that's a transit agreement. Their traffic transit passes through someone else's network and they pay for that. A peering agreement is when two ISPs make some agreement saying, let's exchange our traffic for free. Let's not pay anything and say, okay, usually with two ISPs which are about the same size, they say, I will carry the traffic of your customers if you carry the traffic of my customers. And usually there's no cost involved in that. So we get two general types of agreements, transit and peering. And also we can usually refer to ISPs, internet service providers, based upon how big they are, how much coverage they provide across the globe. And often referred to as three different levels, three different tiers. The top tier, tier one, tier two, and tier three. Where the top tier are the big ones, the big internet service providers, tier two are the medium size and tier three are the small ones. Tier one internet service providers do not pay for transit for any destination in the internet. They either have a network covering the entire globe, unlikely, or what they do is that they have peering agreements with other tier one ISPs. And there are only about between 10 and 15 tier one ISPs in the world. So the big companies like AT&T, Japanese NT&T, and a few others, the list of companies changes over time. There are about 10 to 15 different companies which are called tier one ISPs. They have connections directly with each other tier one ISP. So once your traffic is on one of the tier one ISPs network, it can reach any other destination in the world and for free because they do not pay transit from anyone else. The next level down are tier two ISPs. They are large, say national ISPs or large companies that have coverage across most of the globe. They must pay for transit from some tier one ISPs and they may peer directly with other ISPs. And then at the next level we have the smaller ISPs, tier three. You and I, as the end users, the customers, and also SIT or SIT is part of TU's network, we pay for transit from one of the ISPs. This diagram tries to illustrate that. Instead of internet service providers, instead of ISPs I've drawn autonomous systems here. At the top we have just three, but there's in practice about 15 in the world. These are the tier one ISPs or autonomous systems and they have peering agreements with each other. Once there's data on this network and it needs to go somewhere over here, if this organization doesn't have a direct connection they can send it via one of the other tier one ISPs. At the same level these peer links here refer to peering agreements. They don't pay to carry each other's traffic, it's free, that's the idea at least. Now here am I down the bottom here. I pay, I have my own network for example or it's SIT, some company, we have our own network. We pay some let's say tier three internet service provider for access. I'm a customer of one ISP, let's say of this ISP. So I pay them to carry my data and I want to send it to someone on network E from network A to network E. So I pay my ISP to carry my traffic but to get to network E, my ISP here needs to pay this ISP to carry their traffic as well. And similar, this one pays a tier one ISP to carry the traffic and then we get down to the destination here. So you pay your ISP, your ISP may pay another one. They may not pay on the same terms as we do but that's the arrangement. So that's a transit agreement. I pay my ISP for my traffic to transit their network, to go through their network. My ISP pays theirs to transit the network, to pass through their network. Some of the ISPs may realize that what's happening is that they, let's say, every time I send to network E, AS6 needs to pay AS1 for that traffic. The traffic needs to transit AS1 down to network E. So what may happen is that AS6 or the ISP here with AS4 may come to an agreement and let's say, let's bypass AS1. Let's create a direct link with each other, a peering agreement. So first create a physical connection and then make some agreement and say, okay, everything that comes from AS6 to AS4, let's say in this path, they will send for free, AS4 will send for free and similar, everything that goes from AS4 to AS6, they will send for free. They make some agreement and say, I will carry your traffic if you carry my traffic and hence neither of them need to pay for AS1. So there's the benefit of having this peering agreement. You no longer need to pay for your higher level ISP to transit through their network. So there's some commercial advantage of doing that. So long as they can make some agreement and say, let's carry each other's traffic. It depends upon the size of the company sometimes and how much traffic is likely to carry or pass through their network. So transit agreements, you pay someone else for them to carry your traffic. Peering agreements, there's an agreement to freely exchange traffic between different ISPs. A peering agreement, of course, can make it cheaper for the end customer. Without the peering agreement, AS6 pays AS1. Well, the cost of doing that is eventually passed down to the end customer. With a peering agreement, they no longer have to pay and hopefully the cost or the savings that they make is passed down to the end customer as well. So it can affect the end user as well. It can also give better performance, not just cheaper, but instead of having to pass through some other network, a direct connection could increase performance. So many ISPs start to create peering agreements with others. So in summary, we have physical connections between peers, either via a direct link, a private peering link, or via some internet exchange point, a public peering point, so they're the connections. Then we have agreements between the ISPs. Two types, transit and peering. Transit, you pay someone else for them to carry your traffic. Peering, you exchange traffic freely between two ISPs. And finally, there are broadly classifications of ISPs based upon their size and their network coverage. Tier 1, they do not need to pay anyone else to carry their traffic. They have peering agreements with all other Tier 1 ISPs. So to reach someone over here, we just send to another Tier 1. Tier 2 are the larger ones, but still must transit some Tier 1 ISPs. Tier 3 are usually the smaller ones. There's no definition of what size. It's about arranged by the market. That's a rough explanation of the commercial arrangement and the structure of the internet service providers in the internet. There's no formal definition. Some of the terminology may differ, but I think it gives us enough of a view on that. Last two slides, I think, before we finish this topic, this one's about the internet exchange point. So now there are companies that provide this service. Many ISPs connect in, physical connections, and then they make their transit or peering agreement between the companies. So an internet exchange point. Usually large buildings or large data centers that have high connectivity. They support hundreds of different ISPs to connect together. Last thing, I've always talked about an ISP an internet service provider, but we see now more so that it's not just people selling internet access, but it's also companies that are what we may call content providers that are having their own network on the internet. The companies that many people access content via have their own autonomous system and their own network or connections into ISPs. YouTube, one example, Microsoft and so on, they have their own networks that connect directly into Thailand, for example. If you look on the internet map, you'll see they are one of the blocks in that internet map for Thailand. Why? How it works is that these companies, for example, Google, rather than having all the traffic that goes to the YouTube web server, rather than having to pay some ISP for their traffic to go through their network, they create a direct link bypassing some ISPs and making it cheaper to carry the traffic. So Google, for example, is based in the US. In the past, from Thailand to access their servers, you need to go via the international gateway and then go across to the US and access the servers. Now Google has a direct link into Thailand, I think via Singapore, which means your traffic doesn't pass through many different ISPs. It goes through your ISP, let's say True, and then direct to Google's network. So it doesn't have to pass via different ISPs to Google's network. That's cheaper. Cheaper for Google, because there's less traffic or less they pay other ISPs for, and often faster, because you have a direct physical connection. So good for the customer. So you start to see many content providers start to have their own network and that summarizes what we know about the internet. So what I want you to pick up from this topic is what are autonomous systems? What's the difference between an interior and exterior routing or gateway protocol, BGP versus the others? The concepts of peering between different autonomous systems, the physical connections we can have between them, we can have direct, private peering and shared public peering, and the types of agreements, transit and peering agreements that we may have. The main things to pick up from this topic. Any questions? Any questions about the internet? All quite easy so far. Nothing too technical, nothing about details or protocols. So the question is there a possibility that the tier one ISPs peer with lower levels? By definition, the tier one ISPs peer with every other tier one ISP so if there are three here, they have peers with every other one it is possible to have a peering link with one of the lower level ones from peer AS2 to AS7 it is possible. Yep. And similar, and sorry, and similar between tiers it's possible to peer between them, so from here to here. Depending upon the normally the size and the types of traffic those ISPs carry. Why does tier one not use the internet exchange point? I didn't say that, it could. It can. This peer line here means the agreement. Distinguished between the physical connection and the contract with the agreement as to what they carry. These links here are not physical connections or not necessarily physical links. They are referring to some arrangement. This one pays this one. This one pays this one. That's what these links show. The peer link these exchange freely. How do they connect from here to here? It could be a direct connection, it could be via an internet exchange point. The physical connection may not directly map to what I show on this slide. So the benefit, so currently if let's say you're on network A and you want to send to me on network C then the traffic must go at least via AS6 and down here to C. And that is the traffic from ASA must pass through AS6 and therefore ASA must pay AS6 for that traffic. If these two create a peering agreement between each other and have a physical link between each other then the traffic can go direct through here. ASA no longer pays AS6 the traffic goes direct and is free to go to the destination C. Why would you do that? Cheaper for ASA? Why would you not do that? Well AS9 is only going to do it if they have traffic going the other direction as well. Because it's free in both directions. It means that they must come to an agreement and say okay every month or every day I'm going to have about the same amount of traffic coming into your network as you're going to have coming into my network. But if for example every month there was one megabyte of traffic going from 8 to 9 but just one megabyte going from 9 to 8 there's no benefit for AS9 in that case. They do not save anything so why would they do it? So it comes down to a commercial arrangement. So normally it's about the similar size ISPs that provide similar service. Is there an ISP in Thailand? Categorize as a tier one. Let's check. There's no hard definition watts of tier one but let's see a list. And who has it? Wikipedia has a list. So it has some description but down the bottom it has a list of tier one ISPs. The name where they are, the AS number and some links. So the answer to your question is no you see the list here. AT&T Century link. So US companies German company here Verazon, Sprint from Tel-Isanera, Sweden, Japan and one from India so no. The large ISPs here, true I guess would be considered tier two possible tier two ISPs. National coverage of the nation maybe considered tier two. And then the smaller ISPs maybe just in Bangkok or in some city tier three ISPs. And before we finish for today one more picture or one more example we will not start the next topic. So one of those companies Telefonica an ISP or a company from Spain just as an example this is some pictures of their network let's have to zoom out and zoom in so just a picture of their links that this company has across the globe this is the globe and we will zoom in on some parts but these are the orange ones what they own so they own some cables, optical fibers and so on or satellite links and the white ones they lease or they have some partnership so they may not directly own it so this is one of the tier one ISPs Telefonica and their network so the the lines indicate the links that they have across the globe that's maybe as far as I'll go so for example they have multiple links across the Atlantic all through Europe and see through the Middle East and across to Asia and so on so most of those ISPs or some of them you go to their websites and you'll find similar network maps that show the connectivity that they have and the links they have into Asia, Southeast Asia so links into Thailand and other countries around here so that's one example of a large tier one ISP it's better to look at their website to see sometimes they have some interactive or some flash maps that provide more details so then for example a local Thai ISP may connect into their network so if you want to connect send your data to Europe your data travels through let's say you're on TU's network it goes through Tru's network it goes to the Tru International Gateway and the Tru International Gateway then has a link into Telefonica's network which then may travel to wherever to Indonesia or down to Malaysia and across to India and so on or up to Hong Kong any last questions? everything's okay, you don't need lecture notes you're you remember everything no need to make notes, that's fine okay enough for today tomorrow we'll start on the next topic of Wireless Lands and if you want to look at some of these there's some links on the website or come and see me but it's best to look at the maps in your own computer, you can zoom in and see the details let's continue tomorrow on Wireless Lands okay and the last two minutes