Definitive MPLS Network Designs [Electronic resources] نسخه متنی

Traffic Engineering Within the USCom Network

QoS Design in the Core Network" section, one of the fundamental network design rules adopted by USCom is the overprovisioning of available network resources, hence ensuring bounded link utilization in the core. This implies the following:

Sufficient bandwidth must be provisioned in the core optical network.

The metrics used by IGP (IS-IS in the case of USCom) must be computed so as to efficiently balance the traffic load in the core. In other words, traffic engineering is achieved through manipulating the IGP metrics. To ensure that link utilization remains below 40 percent in the absence of failures at all times, USCom decided to develop an internal tool that computes a set of IS-IS metrics that are used to traffic-engineer the network. The tool is run on a regular basis (approximately every 6 months) to accommodate some traffic growth. It is triggered by the monitoring of link utilization in the network by the USCom management system.

Note

It is worth mentioning that changing the IGP metrics is not a completely cost-free operation. It requires some nonnegligible work for the network operations staff. Indeed, each IS-IS link metric must be changed individually to give the network time to converge before changing another IS-IS metric. Moreover, although transient states may lead to temporary congestion, as already stated in the section devoted to the network QoS design, this is unlikely to impact the overall SLA because it is averaged over a period of one month. Furthermore, such changes are performed during maintenance windows. Another constraint added to the IGP metric computation tool is to keep the link utilization below 70 percent in the case of a single link, node, or SRLG failure. In the case of the USCom network, this was an achievable objective in most cases.