 We enjoy Netflix, IFLIX and Amazon Prime, etc. as entertainment services, but in reality, we don't know, or even if we know, these are the content delivery networks or content distribution networks. In couple of lectures, we are going to look at how content distribution or delivery networks are designed, what are these and we'd look at some of their advanced topics. We'll start off with basic concepts regarding CDNs, their definition, the services and we'd look at state of the art. Content delivery network is basically a collaborative collection of network elements and servers spanning the internet. Sometimes it is also known as OTT service over the top service. The content is replicated over several mirrored servers. These servers could be mirrored according to any RAID standard, for example, you might be knowing RAID 5. So these CDNs perform transparent and effective delivery of content to end users. By transparency, what we mean is that the user is not concerned about the location of the content because the content could be mirrored and the particular delivery would be taking place closest to the user or even farther from the user. And then the effective delivery means certain quality of service parameters have to be provisioned. In order to do that, the CDN components must collaborate with each other in order to have provide a better user experience. The services that we could conceive for CDNs would be request direction and content delivery service. It means the requests from the users have to be directed to the most appropriate CDN server, the original CDN server or its cache server. So using this mechanism actually enables a content delivery network to avoid or bypass congestion. Two well-known examples of congestion are mentioned here. One is the flash crowds where a very well-known website is accessed for millions of customers through hyperlink, which is provided as let's say an ad on a not so well-known limited resource server. So that small-scale server or small-scale web server would be starved or it would be choked because it is relaying the request to that well-known web server which has a lot of resources at its end. Then we have something known as slash dot. Slash dot is a very well-known, highly reputable magazine that publishes content related to the advancements in technology. So there's something known as slash dot effect. What it means is the users or the frequent visitors to slash dot website would get to know of an interesting application or an interesting new service because slash dot is enlisting it or is mentioning it. So that would result into an overwhelming response on the particular web service that poses a lot of traffic overhead to it. So it means request redirection to bypass congestion is a fundamental requirement for CDN. Then we have the content outsourcing or mirroring or caching and distribution. So this involves replication of and caching of content from the original server to the other web servers. Then we have the content negotiation. If you recall back in old days we had something known as a service-oriented architecture in which we had user agents, service agent and directory agent. So the content negotiation service is essentially what allows the users to have very specific individualistic experience by getting the content that meets its demand the most. Then we have the management services which are required. For instance, we need to manage the health and diagnostics of the network components. We need to make sure we have triple A kind of diameter or radius services available. And then the content usage has to be monitored and reported for better service provisioning. Let's look at how typical content delivery networks are involved in providing contents starting from streaming media to web pages to mp3 content. Then certain cell phone services like WhatsApp, we have remote virtual desktops, we have thin clients like laptops, etc. All these contents and services are provided by different kinds of CDNs. The architectural components of CDN at very basic form include we've got users. Then we have the user making request to the request routing system which in turn is going to refer to the user to a replication server. The replication server is nothing but a mirrored outsourced site for the original content coming from the origin server via a distribution system. And of course, we need a billing and accounting mechanism. So if you really want to understand the existing operation and comparative performance of different kinds of CDNs available. This is a very nice paper. It is actually a study on the active measurement for two video streaming services like Netflix and Hulu. What the authors have done is they have taken a range of key aspects of OTT service platforms and then they have compared these against the streaming protocol. For instance, Netflix uses Dash and Hulu uses real time messaging protocol. And then the CDN deployment strategy, CDN selection strategy and the user experience. So the authors have compared these Netflix and Hulu. Here high level architectures are shown here in a typical Netflix situation. We've got the user making a request to the data center that subsequently makes a request to either the cloud sourced environment or is right outsourced to a dedicated CDN that provides Netflix like services. So this is the general architecture of how Netflix works. Of course, Netflix has its own proprietary hardware and software and protocols through which it gives such a smooth user experience. For Hulu, we've got a more numbered visualization. We've got the page request and then the manifest file is obtained that contains reference to the CDNs CDN servers where the content possibly could be deployed. And then the reports are being generated on the usage of certain content. The papers that I've referred are two. The first one is the lecture notes on an electrical engineering. It's a beautiful book by Raj Kumar Bhaiya, Mukhtar Prathan, Mahdoom Prathan and Vel Vaikeli. This was published back in 2008. The paper that I have referenced to you might as well like to have a look at it is measurement study of Netflix, Hulu and a tale of three CDNs. It's a very prestigious journal IEEE Transactions on Networking back in 2014.