 Content delivery networks, as we know, have the potential to accommodate and provision an arbitrary, arbitrarily large number of users from anywhere in the world. There's a problem associated with limited infrastructure when it comes to the content delivery networks because they have to rely on support from other content delivery networks or other data centers within a content delivery network if it is overwhelmed. This is the issue or the problem of request redirection. The problem gets more complicated with the presence of the dynamic content, which is different from static contents as we know, which are simple files which can't be altered. In this module, we are going to look at the Dianal effect, the dynamic content and its relationship with the request redirection. Dianal effect, as we know, is the time of the day effect, which is going to have a peculiar impact on the infrastructure of content delivery networks when the users become increasingly high. This is observed almost at every website. The problem is even further complicated because of flash crowds. The typical example given is of Victoria's. This situation can cause a significant delay in the service time, so much so that up to 20% of more load could be observed and corresponding 20% more time in the service time. How could this be handled? It could be handled by provisioning resources on certain estimates. If the estimates are based on peak utilization, then we can expect better response time, but this is going to come at trade-off because the system is largely going to remain underutilized because not always the system is peakly utilized. Similarly, on the other side, if we now measure the load as a mean, in this case, the dimensioning of the network on the mean load would result into better network utilization, but it would become problematic whenever the network traffic surges. Let's look at the problem complication when we have dynamic content. These days, each user has its own profile and has a personalized view such as using cookies and by login. So it means if a user wants to have a more personalized experience, it needs to have content that is dynamically generated, you can say on the fly. This could depend upon the user's profile, user's location, the past history, and of course the particular nature of the query itself. So the content in the form of a web page would now comprise the static content and the dynamic content. The dynamic content is different from the static content in that it burdens the infrastructure in a certain way. For example, if it requires a new process to be created, for example, using CGI scripts, then some kind of compute burden is going to arise. Similarly, if it is accessing database, then particular type of database query is going to have a certain kind of PHP script which would in turn execute. So it means that for dynamic content, it is going to burden the compute infrastructure. Whereas for static content, the requirement for looking up at the load balancing strategies or service optimization strategies, the typical requirement would have been to look for the nearest server. But in the dynamic case, it's not always that situation. So here, in addition to being the closest server, the server also has to be least loaded. Let's now look at the request redirection that is the request is from the client is now going to be redirected to the corresponding server. Depending upon which particular part or the tier of the network has to be load balanced or redirection is to be executed at what point. So we have the user that initiates query or client request. This particular request goes to the to the access tier. And then on the back end, we have typically the web server as the web tier, then the web server could be getting dynamic content that is getting executed runtime at the application tier, and the application tier could in turn be connected to a database tier. So redirection is now going to be either initiated by the client itself or by the server. The request initialization at the client or server is has its roots in making a decision whether the client is going to handle it, its redirection itself, or it's going to be the server doing that. So if it is the client, then the client query is redirected by the client itself to a server by a proxy which is installed on the client side. Akamai networks, which is one of the largest and where most well known content delivery networks in the US, this kind of mechanism could work. The client side proxy could also be termed as the server or an entire overlay. The client side is going to make a judgment call on the basis of network traces and measurements, for instance, by using simple tools like trace root or ping. Then the client actually selects the server in the network proximity which is going to serve this request best and is also least loaded. Now this particular mechanism is known as URL rewriting. This also is going to involve DNS name to IP address resolution. The server side redirection is more robust that the client does not have to do much. The request is forwarded to an initial cluster or a certain data center. There the redirector service within the cluster or the data center redirects the request away to the best server. Let's look at the redirection example. We already have understood the multi-tiered cluster architecture. On the basis of the same architecture, if we look at the figure on the right hand side, we see that we have an initial client request arriving at the Internet data center on the left hand side. Here the redirection is going to be activated not at the web tier, not at the application tier, but it's going to be redirected to a different database tier. It means probably the data center is at another place which hosts the right database with the right resource records. So the redirector service dispatches this request or forward the request to the most appropriate Internet data center. Now this kind of forwarding or redirection serves the purpose and it helps the fastest provisioning of dynamic content for the CDN users which are looking at a certain content for application provisioning. We are continuing with the same book of Raj Kumar Bhaiya, Prathan and Vakali. You can actually look at I guess it's chapter number five or six.