* Revisiting designing MPLS nets with Cisco routers
In a recent newsletter entitled “Designing MPLS networks with Cisco routers” we discussed how service providers deal with MPLS service classes somewhat differently depending on the equipment vendor they use. In response to that newsletter, we heard from Azhar Sayeed, manager of product marketing at Cisco who wanted to clarify how Cisco equipment is used to implement the queues that correspond to the MPLS service classes.
Sayeed pointed out that packet buffering and scheduling only occurs when there is congestion. He stated that Cisco equipment supports one high priority queue as well as a number of lower priority queues, and that the vast majority of enterprises use a total of four or five queues (service classes). He also stated that if there is any traffic in the high priority queue, that traffic is transmitted before the traffic in any of the other queues is transmitted.
For the sake of example, assume that a hypothetical company has a T-1 access link and that it has allocated 40% of the capacity of that link for high priority traffic, which is typically voice traffic. When there is congestion on the T-1 access link, the traffic is policed and any high priority traffic in excess of 40% of the capacity of the T-1 link is dropped. This is done in the belief that it is better to drop an occasional voice packet than it is to deliver it late. Alternatively, if there is no other traffic on the link, the entire capacity of the T-1 could be used to support high priority traffic.
Independent of the equipment vendor that the service provider uses to provide MPLS services, IT organizations have a challenge managing MPLS service classes. One part of the challenge is to ensure that the company’s applications are mapped to the correct service class. Another part of the challenge is to do capacity planning for each one of the MPLS service classes. We will explore this challenge in detail in future newsletters.
