 Let's look at the quality of service-based actual services which can be provided in Ethernet-based NGNs. For that, we'll have to consider the importance and the dominant role of the operator, then the inter-operator rules known as the edge rules, then we are going to look at the services specifically and then the traffic classes. We've covered traffic classes already but we are going to recap these. So all the QS services are actually operator-based. It means a user belonging to a certain operator may request QS-based service provisioning from the native operator. The operator is going to provide first and foremost the connection which is based on some kind of admission policy. It is known as the connection admission control mechanism. Through this connection admission control mechanism, the user request may be denied, may be serviced, may be downgraded or anything else can also be possible. Then we have the parameters which are also set or adjusted by the operator. So the operator beyond doubt is the most dominating player in this thing, of course, after the user because the user actually has some kind of service level agreement with the provider. So the UNI and NNI parameters are adjusted according to the operator scope and provisioning. Then we have the concept of application of rules between the network, two different networks or at the network boundary. So the edge network nodes are the ones which lie between the access and core boundary. So these nodes are very critical because when homogeneous network is there, then this quality of service provisioning is an issue that is homogenously or seamlessly handled. But once we are crossing the boundaries, then the edge network nodes have to have a role to play. So these edge nodes apply certain rules which are certainly through the understanding and mutual consensus of both the access and core network operators. So these edge rules actually are based on traffic treatment from measuring the traffic to regulating the traffic to just marking the traffic or in the worst case, dropping traffic altogether. The QS services, which are based on these functional aspects we've just discussed, can be categorized into an exhaustive list. But from the brevity point of view, the QS service types for Ethernet are regarded as best effort being the simplest and the lowest in quality followed by gold and premium being the best of all. These are categorized or classified on the basis of the traffic parameters which have their own quality related attributes, the edge rules which are applied at the boundary, the bandwidth allocation or the bandwidth profile that we discussed and to which particular QS class of the ITU does this service type belong to. So we have the premium service for instance, which has constant bit rate. It has a committed burst size that is a burst is a sudden upsurge of traffic which is certainly a deviation from a constant bit rate. So if this burst size is agreed to, then it means it's a committed burst size. Then no excess bit rate or bursts are granted in the case of premium service. The edge rules are actually very stringent here because only then premium service can be provided that any flow or comprising Ethernet frames which are not conforming to the CBR guidelines or the committed burst size guidelines shall be straight away dropped. The bandwidth in premium is therefore pretty dedicated. The QS classes for IP traffic by ITU can be class 0, 1, 6, and 7. We've discussed all these but we are going to recap them shortly. So you can have a look at other service types as well. The only thing which is different and most noticeable between premium and gold is that the excess bit rate and burst sizes are allowed because this is statistical time division multiplexing in which inactive bandwidth users give away voluntarily their bandwidth to the active users. So the active users can actually consume or occupy the bandwidth not used by the inactive users. Then we have best effort which is based on class 5 and in best effort everything is at the lower side. This is a recap of the IP traffic classes which were defined by the ITU. Just for for gravity, let's look at class 0. The upper limitation on the delay variation is not more than 100 milliseconds. Then we have the transfer delay is not more than 100 milliseconds. The delay variation is not more than 50 milliseconds. It is quite specific and then packet loss ratio is not more than 10 to the power minus 3. That is not more than a single packet in a thousand packets can go in error. Then we have the packet loss and packet error ratios which are again very stringent. We've already covered these if you recall we had class 0 and 1 for real-time jitter sensitive applications which is the most crucial kind of traffic that has to be serviced as a multimedia traffic over NGN. Then we have class 2 and 3 for signaling and interactive traffic and subsequently class 4 for video streaming.