 Let's look at the flow state aware transport once again for QS provisioning. In order to implement the FSA based transport, which complies to certain QS requirements, there are going to be a bunch of network elements that we have briefly seen, protocols, and some interesting algorithms which are all going to work together to realize the transport of the flow state aware traffic. Let's start with the flow identity followed by the bandwidth management and we just recap the flow aggregate that we already know. The flow identity is basically the identification of an individual flow. The support is available in IP networks, both for IP version 4 and IP version 6. Specifically, IP version 6 provides flow label. It's an octet available in the IP header. It is joined together or it is concatenated with the source and destination IP addresses which can help this flow to be identified and this is then stored in the flow aggregate table. For IP version 4 based networks, which is the dominant case anyways on the internet, the IP addresses and the port numbers are utilized. For voice over IP, multimedia traffic and IP TV using session initiation protocol set of or suite of services like SCTP, there is an identification of the endpoint through which a flow is identified. Then we have the next important thing that is what possible algorithms can be implemented to perform bandwidth management. It is again a very important requirement for QS provisioning. So, bandwidth allocation is basically a policy issue that is who gets to have what bandwidth. This is implemented predominantly through explicit means or through implicit means. Through explicit means, we have the well-known scheduling disciplines such as FIFO for best effort. FIFO actually means no specialized priority. Whatever comes in or reaches the input buffer of a router gets corresponding treatment accordingly. But if you want to provide QS for multimedia traffic, then weighted fear queuing is utilized where weighted coefficients are assigned to different flows which in turn are going to be treated differently by the input queue of the router or a switch. Well, this is what we call the more explicit kind of bandwidth management by the network elements. For more implicit or desic kind of bandwidth management, we have priority scheduling. When we say why it is implicit is because of the fact that when we talk about priority, a priority is a relative term. So, it means when multiple competing flows are going to be there, then one of them would have a higher priority over the other. But if there is no competition as such or if the flows actually belong to the same priority, then this bandwidth allocation becomes more of a best effort kind of response. Then we have the flow aggregate. We already know that flow aggregate is a bundle of flows. These flows actually should correspond to the same QS requirements. So, it means the flow aggregate has to be one for certain kind of flows with the same QS requirements and has to be other for a differing bundle of flows. Now, each flow aggregate basically is identified by several parameters. The most obvious one is the flow aggregate identifier that is a label switched path. It's a label which is assigned to a certain flow in MPLS networks. This is basically the most obvious one on the basis of this parameter value. The flows can be treated. But more interesting features which are of concern to network designers are the first one being the label switched path itself. Then the actual number of flows which are going to be in a given flow aggregate because is it going to be an indefinitely large aggregate or it is going to be limited. Then the ratio of the maximum packet length and minimum link capacity. This is a very important consideration to form the flow aggregate because sometimes there are certain flows within a flow aggregate which would be sending traffic with very long packets and there would be some which would have very short packets and average packets. But the main concern or the trouble comes from the maximum length packets and this problem becomes very obvious when this flow aggregate has to traverse across a number of intermediate networks. Each network would have its own link capacity. So an interesting feature of the flow aggregate is going to be a ratio of the maximum packet length and the minimum link capacity. This is going to be calculated for all the links. Then the number of hops within an aggregation region that is it is the intra region hop. Then the maximum number of aggregation regions which would be connected to provide an end to end flow or to handle the flow aggregates. So it is important to consider these while defining the scope of the flow aggregate. Now mapping of a flow ID an individual flow ID to a flow aggregate is again dependent on certain other interesting parameters that is the maximum sum of sustainable transfer rates for all flows is going to be very important. That is what is going to be the resultant size of the flow aggregate after including all the flows and each flow is going to operate at its peak. So it means this flow aggregate is going to be a heavy flow aggregate then likewise maximum sum of sustainable burst sizes. Now there is an average maximum transfer rate for every flow then each flow can also have a burst temporarily. So is this flow aggregate going to sustain that are the network elements going to accommodate such a flow aggregate into their network. Then what is the discard policy and what is the scheduling policy of the service providers and what are the service classes which are offered by the service provider because the service provider may be providing only some or in an ideal case all of the services which the flow aggregate contains. The operator policies are eventually going to be the final say or the final word of course each individual operator is also connected to other operators. So an inter operator agreement is also going to have a final impact on how exactly an individual flow can be mapped on into the flow aggregate. So this is again the same figure that we have seen the flow aggregate information has to be shared or disseminated between the network elements through the information exchange functional entity. Otherwise when flow aggregate is finally admitted into the network and if the network elements are not correspondingly geared to achieve the quality of service for the flow aggregate then some network degraded performance would be experienced.