Module 3 Part 1

Hello and welcome to the first part of the network architecture lecture series. In this video, we are going to look at why network architecture is important, the components that are used to create a network architecture. We'll examine the idea of backboard networks, or enterprise networks. Which are used to create networks that connect all the different parts of an enterprise together in one physical location. Then we will look at the idea of virtual networks, or virtual local area networks. Which are a way of creating networks that are independent of the physical wiring or the physical layout of the network itself. You could think about it as a network that can be reconfigured using software. Before we get started, I wanted to say that this material does require a little bit of background knowledge on TCP IP and the OSI models. For those of you that don't have this background, I will go over that in the life class. I've provided a couple of reference documents. The overview of TCPIP is one that is easy to read and understand. I also have the IBM red book. And I must warn you that it is well over 1,000 pages. I just wanted to have it out there, since that is considered one of the definitive references on TCPIP. At an extreme level of depth, network architecture combines a number of different technology solutions, such as networking equipment, cabling, wireless technologies, to a complex environment that provides for highly available. That is, it is available all the time. It is secure. It is scalable that it can grow with the demand of the organization. It can be managed and it is reliable. When we look at network architecture, there are two different objectives. One is to provide reliable and efficient movement of data packets from a source to a destination. Goal number one is, let's get data in the form of data packets from A to B. Where A and B are two different systems, two different network devices. But the big thing that we want to introduce at this point is that, that movement of data should only happen if it complies with organizational security policy. When you take these two objectives together, it does seem somewhat contradictory, right? The first goal is to move the data from one place to another in the most reliable and efficient manner. While the second principle is to restrict the types of traffic that can flow from one device to another. However, keep in mind that these two principles are really closely tied together when we design a network architecture. Before we can dive deeper into network architecture, we need to understand the three basic types of networks. This categorization of networks is really based on the geographic span. Basically how big the networks are. Local area networks are the networks that are used in small buildings or maybe a floor of a larger building or a portion of a floor of a larger building. These could span about 100 meters, 330 feet or so. That's not a distance that is set in stone. Metropolitan area networks are those that connect entities that are in a larger metro area, so to speak, the Greater Chicago area or the greater New York City area. These could be a few miles in terms of the size or the span of the network. Wide area networks could be pretty much anywhere in the world. When you think about an enterprise that has got multiple locations worldwide, and all of these locations are connected together. The connection between those networks, what creates a wide area network? There is really no limitation on the span of that network. In addition to that, there are a couple of things I want to stress here. The term internetwork basically points to interconnected networks. Or a network of networks, right? This idea of a network of networks, when taken to the largest extent possible, creates what we call the Internet. And I'm calling it the Internet because it is a single entity that is made up of a large number of individual networks that are connected together into the Internet or the Inter network. The Internet is the largest example of an inter network. It's basically a network of networks that spans the entire globe. Or I guess when you throw in things like the International Space Station, it is basically the largest network that human beings have built. Now one of the things about the Internet that's important to understand is the Internet is not collectively owned by any entity. Each of the networks that make up the Internet are owned by different entities. The connectivity between these networks are owned by entities as well, and these entities typically tend to be the Internet Service Providers. They basically connect all these networks together. The ISP's or the Internet Service Providers do own portions of the Internet, but nobody owns the entire Internet. Now let's take a look at the concept of a backbone network. I mentioned that a local area network, or a land, has a span of about 100 meters, or a few hundred feet. If you think about a large office building or a small canvas that needs to be made up of a number of these local area networks together. Backboard high speed networks, that basically connects the local area networks together. And then it could connect to other backboard networks. It could connect to metropolitan area networks. It could connect to wide area networks. Or it could connect to the Internet. When you think about Indiana University and our Bloomington campus, which is where I am, we could have a backbone that connects all of our buildings together. And that's what we call the campus networks. If you think about the Kelly building, we have multiple local area networks connected together. The same thing in the library building, the same thing over at the Luddy school, the same thing over at the psychology building, next door, the geology building, the IMU. Then we have a campus network that connects these fragmented backbones together into an integrated backboard network for all of Indiana University Bloomington. The idea that a backbone network is also an inter network and is often referred to as an enterprise network. When we call about an enterprise network or when we refer to an enterprise networks, that basically means that the local area networks that are involved are part of the same organization. Now, wide area networks, which are the connectivity between far flung locations of an enterprise. Those wide area networks could be provided by a third party service provider, much like an Internet service provider. The technology for these backboard networks, at this point, most of the competing technologies for backboard networks have died out. Ethernet, which is basically what we use at the local area network level, is also now the dominant technology for the backboard on the backbone Ethernet network is capable of running at one gigabit per second, ten gigabit per second, or even 40 gigabit per second Ethernet. Now, when you get to the higher capacities than we typically run it over fiber optic cable, and because it's fiber, some of the distance constraints are removed. Right? So the hundred meter constraint that I gave you earlier for a local India network is basically based on copper wiring, electrical signals. When you go to fiber optic wiring, that it could actually go much, much further. Right? With that said, what you find on a backbone network would be very high speed backboard connectivity in the data center, which is where all of the systems that are shared, like the servers, right, The application servers, the web servers, the database servers, all of those could be connected together into a 40 gigabit per second. Ethernet buildings might have connectivity that is running at ten gigabit per second. And that within a building it might be just one gigabit per second, right? And understand that one gigabit per second is tremendous amounts of data and most computers that are in use today would max out their capacity when you are feeding it data at that speed. The backbone is connected together using one of two different types of devices. This could be a switch or a router. A switch basically connects computers together. While networks together switches connect computers into a network, into a local area network. Then a router can connect multiple local area networks together. A switch comes in two different versions. A layer two switch uses the data link layer to move data or the physical address is used to move data. Layer three switch uses the IP address. Now, in today's network on smaller networks, we will find layer two switches. On larger networks, we will find layer three switches. Basically, for this class, I really do want to limit the discussion to that level and not get into the intricacies of network design and switch design. Routers connect multiple networks. Now, a network is connected to another network that is owned by a different entity. The router that is used is called a gateway. All gateways are routers, while all routers are not gateway. Basically, a router is a P based device and Internet protocol based device that is capable of connecting multiple networks together. And basically routes the data. That is, when data comes to the router, it decides what is the router or the next best hop, if you will, to send it to a gateway. Connects networks that are owned by different entities. An enterprise networks connection to the Internet would be through a gateway. An enterprise networks connection to the wide area network would be through a gateway. So when we look at the backbone network types, broadly speaking, the most common type that we see is a routed or a layer three backbone. So this basically moves data along the backbone on the basis of the network layer address, which is the IP address, right? And the connectivity is using either net. Usually in a modern network it would be a gigabit per second Ethernet. If the building has not been wired lately it might be 100 megabits per second, right? So this is commonly referred to as a subnetted backbone because every network is basically a subnet of the larger Internetwork. Now, when we use a routed backbone, the membership of a computer to a network is basically based on the switch to which it is connected, right? So the physical connectivity of the device to a specific switch determines its network membership. A virtual lane changes that. In a virtual lane, the network to which the computers are assigned is done by software rather than by hardware. And this mechanism where we are dissociating the underlying physical layout of the network from the logical flow of data through that network is becoming very popular. This actually allows somebody sitting in, let's say, building one of your canvas to be part of a local area network that is over in building five. Although they are physically in two different networks, or the sub network membership can be determined by using software, diving in deeper a routed backboard. Every local area network is a separate and isolated network, which is connected to other networks through a router, to some kind of a backboard network. Within the local area network, the messages are sent based on the data link layer address. And then the data that needs to go outside of that sub network is sent to the router, which then forwards it to the network to which it is supposed to go. If we look at this routed network, basically each of these are separate. Local area networks, right? You have multiple computers connected to a switch. That switch is connected to a router and you will see that all of the routers are connected together. That basically is your backboard network. Now the challenge here is that because every packet has to be processed at a router before it can be moved on, that do impose some time delays, especially if you don't use enterprise class networking hardware. This, of course, requires quite a bit of management than just using switches. Layer two switch, virtual lands can be used at the local area level and the backboard network level, They use intelligent high speed switches. Now, the big difference here is instead of connecting physical local area network segments, virtual lands assigned computers to land segments using software, There are two standards around this. There is the triple, which is the Institute for Electrical and Electronics Engineers. Eight oh two, point 1.1 There are two different designs that the standard specifies. You could have a single virtual lane or a simple virtual land using a single switch, or you could have a much more complex virtual land using multiple switches. Obviously, with most enterprise networks, we'd be using a multi switch virtual land. When you look at the types of virtual lands, it could be based on port, which is using the physical port number on the virtual land switch to assign computers to land segments. This basically looks and works just like a traditional wired local area network. We could have a layer two virtual land, where we use the Mac or Medium Access Control address. The layer to TCP IP address, which is usually called the physical address. To form Virtual Lands. Then we could have IP Base Lands, or layer three Virtual Lands, right here. We use the network layer address to form virtual land. Here it is the IP address. Or we could have application base lands, which are called layer four lands because it's a combination of layer four, which is transport, and layer five, which is application. It uses a combination of the port number in the TCP packet, which is at layer four, and then it uses the IP address. Here it is, Port plus IP combination is used to assign systems, which means that at an enterprise level, you could take, for example, Microsoft Edge, which is a web browser, and put all of the browsers into a common virtual land. You could take Oracle, EBS, the enterprise business system, the ERP system, the client software for that could be grouped together into a virtual lane. The advantage of doing this, or even we could use this inside a browser to look at how different sites when they're access through a browser are different applications that access through a browser can be grouped together into different ports. The advantage here is we can actually control the traffic based on the type of traffic or based on the application that is requesting the traffic. You could take your enterprise application and make sure that it has priority over common web browsing tasks. The standards that I mentioned earlier, eight oh 2.1 here, a 16 byte, which is 16 into eight bits, is actually inserted into the 82.3 packet. This is a basic Ethernet packet, and an additional frame is added. Or a tag, or what a tag is. In social media, you're basically tagging a packet as belonging to a certain virtual land. When the packet reaches its destination switch, the land tag is stripped off and then it is treated just like a regular Ethernet packet. Now, multi switch lands use 2.1 P. Here we need to have an additional priority field, you can actually prioritize traffic, right, instead of treating all packets as equal. 82.1 includes the capability of eight oh 2.1 which is the land tag. In addition, it adds a special priority field that tells all the devices on the network the priority that should be assigned to that packet. Taking this, when we think about backbone architecture, it's easy to understand it. You saying a layered approach, right back to the layered approach that we've applied to software. Now let's think about it in terms of architectures. The first layer is called the access layer. This is the technology that is used in local area networks. I'd like to associate the word access by thinking about this is the layer at which all the different devices on a network actually access the network. This is their first level of connectivity to the network, that is a laptop. Whether that is a desktop, that is a server, application servers, web servers, database servers, whatever the access layer connect those computers to the actual network, or it connects the actual devices to the first level of network. The distribution layer is the part of a backbone that connects local area networks together. If you think about the Kelly School building, we have two different buildings, one of which is four floors, the other is seven floors. Each of these floors have at least one local area network connecting all of these local area networks, which is basically the access layer. Connecting all of them together into a building. Network is the distribution layer. That's the part of the backbone that connects multiple local area networks together within a building or depending on the size of a building of the buildings, it could be a small cluster of buildings. Finally, we have the core layer. And the core layer basically connects all the distribution layers together. On the IU Bloomington campus, we have I believe, five core routers for a 2000 acre campus, right, or nearly 2000 acre campus. Clusters of buildings are connected together using the distribution layer. And then all of these distribution layers are connected together to the core layer. Then it is a core layer that is connected to the outside world as the Internet. Right? So in our case, we don't use an Internet service provider, so we directly connect our core network on the Bloomington campus to the Internet access, or the Internet access point in Chicago. Right, that's our network point of presence. This diagram basically illustrates what I just spoke about. Right? We have a multi story building here, for simplicity sake, we are saying that each floor of the building has a local area network that is connected to a switch. This is likely a layer three switch. Right? And then all of the layer three switches are connected together to form a distribution layer. All of the buildings. This is a three building cluster, right? They will all have distribution layers here with one core layer router that connects these buildings together. The core layer basically is formed by connecting a core router which then connects to, you look at this core router, that's the one that connects all these distribution switches together. Collectively, that would be our backboard network. It has got the access layer distribution, then we have the core. If we now want to think about this and saying, okay, let's look at how this is set up in a building. Individual computers all connect to the access layer. Individual devices connect to the access layer in a smaller network, that's typically a layer two switch. Right? If it's a very big network, then it could be a layer three switch as well. In network set of layer three switches connect the layer two switches together, then the core layer in most large networks would be layer three switches. Right now, we could use a layer two if the number of core switches are less because you don't need to really do a lot of routing. As in you don't need to use intelligent traffic management because you don't have a large number of core routers. But nowadays, the trend is to just use layer three switching. Part of the reason why a lot of organizations used layer two switches was that they were cheaper. The price difference has come down in the interest of having a more uniform set of networking devices, Many organizations now just go with a single type of switch, at least for the access layer and the distribution layer. To sum up, networks were initially designed for the fast and reliable movement of data. Nowadays we want to have that, but we want to make sure that the movement of data happens only in compliance with security policy. Our future architecture discussion will look at how do we actually architect a network so that it is inherently secure. We talked about backboard networks. Backboard networks connect multiple networks together to form an enterprise network. It has three layers. Access layer, which is what gives access to the network to all the devices. A distribution layer that connects all the access layer networks together or connects the local area networks together. And a core layer, which connects the distribution layers together and then could potentially to the external world or to an external network.