Add H-CoS for Subscriber Access
People just can’t seem to let a DiffServ network rest. Now that you have IP CoS up and running, the word is that you have to extend this CoS into a subscriber access network that is expected to scale beyond 1,000 users. The design has to be scalable (obviously) and has to offer not only IP DiffServ-based CoS, but also several performance profiles that are needed to enable triple-play services and to facilitate a tiered service offering.
In this example, different levels of performance are achieved through varying IFL shaping rates and level 2 scheduler node overbooking ratios. Clearly, users that are on IFLs with a higher shaping rate with little to no overbooking can expect better performance than those that have to contend for a overbooked group access rate while still being limited by a lower IFL speed. As noted previously, an alternative and safer design option is to only overbook PIR rates.
The design must offer per IFL/VLAN CoS profiles with eight queues per user. So at a minimum, per unit scheduling is needed. In addition, the network architects have asked that there be an interface-level usage limit on high-priority traffic. In like fashion, they also want each subscriber access class (business versus basic) to have an overall usage cap placed the same high-priority traffic. This high-priority traffic is special for having a high scheduling priority, and because it’s often used to carry loss and delay sensitive real-time traffic as well as network ...