 Continuing with our understanding of quality of service provisioning in the core side as well as the XS side, we would now look at a very interesting implementation aspect of QS provisioning known as the flow state aware transport. As the name implies we are going to talk about flows, how the flows are understood, how the flows are aggregated and how the flow state aware transport is finally provided. We will first understand what is a flow followed by we look at some of the network functions which are expected to provide flow awareness and then we look at the architecture which is a standardized architecture for NGN flow aware QS provisioning. As we know that a flow is an end-to-end connection or a communication between two end points which is identified through the source port, source IP, destination port, destination IP and other information depending upon what kind of application is being used. Since we are talking about multimedia each QS multimedia requirement would necessitate that a flow has to be treated on priority. The treatment has to be such that from the initiation of the flow to the termination of the flow all the network elements and all the procedures are within the QS service level agreement. The flow is basically an identification of the quality of service through a class mapping whichever class it belongs to the flow belongs to. Corresponding traffic engineering has to be activated. This activation of traffic engineering can be preconfigured as in static where the system administrator or the network management performs one time QS identification and flow based traffic engineering and then it is repeated for all the flows which comply to this class. Then we have dynamic flow aware transport in which the traffic is based on a flow that corresponds to a certain class. The class and the traffic engineering belonging to that class are in real time brought into the enforcement entities. If you remember we had the functional entities which would enforce the quality of service at runtime. So naturally the static provisioning is going to be limited and the dynamic provisioning is going to be more robust and it is scalable. The QS for each flow is related in terms of the packet delay that a network element such as a switch or a router can provide. The loss ratio actually means the number of packets sent and the total number of packets successfully received. Some of the packets might as well get lost. The bandwidth requirement and of course giving it a priority. So the flow has a specific bandwidth requirement and it also may have a certain priority requirement in case congestion emerges on the network. The network function which corresponds to the flow state aware transport of each flow is implemented in the network nodes. These network nodes can be switches, routers, gateways and so forth. Then the network function also includes admission. We talk about it quite often. We have discussed it in detail. The resource and admission control function is what implements it to either accept or reject a connection. But now in this case remember that we are talking about connection as a flow. The function for signaling are therefore required to reserve the resources for QS support per flow. Now another dimension of a flow is that a continuum or a sequence of IP packets which move from one source to the destination need some consistent and coherent response from the network. So it means that once the QS has been agreed to then the flow as in the total number of packets need the same kind of response from the network. Let's look at in a very basic and broad sense the functions for QS support which are provided by the network. Starting from the customer premises, the customer premises may or may not have QS aware flow state based provisioning mechanism. So we see that we have a non flow state aware customer premises network which in turn is connected to the end system flow state aware functionality. This can be an application which can be installed on the customer premises network. It could be a middleware either hosted on the customer premises or on the network side. Now we see we have certain entities like signaling function QS function which are going to be connected to the service control function resource and admission control function through the service control function. Now the user to network interface is providing connectivity also to the network element or a node which can do in band processing of the functions in band signaling actually is something that we'll shortly talk about and then QS functions. Then we see that this flow which is emerging from the non FSA customer premises network is entering into an aggregation endpoint function. So it means some kind of aggregation is taking place which is accumulating all the flows and creating a flow aggregate. Now this flow aggregate is big fat pipe which is again passing through the FSA nodes routers and switches to perform the same in band network processing for signaling and QS functions. Now this diagram actually tells us that we have customer premises equipment that is generating a flow. This flow is being treated for QS provisioning by signaling. This signaling is in band signaling and then we have the aggregation of this flow with so many other flows and this flow aggregate is then again passing through the network elements. We also need to look at the functional entities which would implement this flow aware transport. So the first one is known as FIM FE or the flow aware information management. Now the information management deals with the kind of information of every flow. The flow information would include the flow identifier in the form of a table. Now each flow identifier would be aggregated or bundled with other flows in the form of a flow aggregate. So it means the flow aggregate would also have an ID which in turn would contain multiple flows each with its own ID. The management of this particular information is carried out by the FIM FE. Then we have the exchange functional entity or the FE FE. Now the functional information, a flow aware information exchange functional entity actually allows this information to be exchanged between different network elements. So it means the flow aggregate table is being exchanged between multiple routers and switches and gateways to develop a good understanding of how many flows are being aggregated in the overall network. Then we have FIG FE or the flow aware information gathering functional entity. This is the entity which accumulates or gathers the information from the entire network or in other words an administrative domain. This handles the requests which are coming from applications because the final decision to treat a certain flow has to get the approval of FIG FE. So we can conclude that the location of FIM FE and FIG are going to be at the respective positions. Since the management is going to take place at the lowest level in the hierarchy where each flow is bundled into an aggregate. So FIM is located in the edge switches. Multiple FIMs can actually be aggregating the traffic in a domain. Then we need to have at least one FE that is one exchange functional entity which is going to allow the flow aggregates to be shared between multiple FIMs. Now this information exchange functional entity needs to be either hosted on a separate server. It can be or it can even be hosted on the same edge switch because if the scope of the network is small then the exchange of traffic is not going to require specialized resources. And then FIG FE is going to be the final entity so it is at the top of the hierarchical network elements and the functional entities. So it is a separate entity placed as a centralized server. This is what we are going to see here. Now this diagram shows us multiple pieces of information all joined together. We have on the left hand side we have an end host that is generating a flow for the other end host on the right hand side. Then we have the FIMs, we have the fees and then we have FIGs. So you see that FIM is the one that receives this flow and then creates the flow aggregate. Now this flow aggregate as you can see is represented by a relatively big long fat pipe shown to you here in white with dotted line running in the middle. This group or bundle of flows moves from one network to the other. So then this information is being shared between multiple FIMs through fees which is at a relatively higher administrative or hierarchical position. Then we have at the top we have the information gathering functional entity which in turn is connected to the application server. So it means the applications would require certain approval, certain admission, certain service. All this thing is managed through the FIG FE.