The link capacity planning scenario is straightforward, because the MIBs such as the IF-MIB and the CISCO-CLASS-BASED-QOS offer all required counters. In contrast, network-wide capacity planning requires generating the core traffic matrix, for which some mechanisms were described in this chapter. The core traffic matrix can be generated with BGP Policy Accounting and NetFlow, especially since the implementation of Flexible NetFlow. Indeed, Flexible NetFlow offers configuration of flexible key fields, which allows any level of granularity. Note that a generation of the core traffic matrix using only BGP next hop as the NetFlow key field offers a simple core traffic matrix and reduces the number of flows to a minimum. The only accuracy approximation with the NetFlow-based solutions stems from packet sampling, which is almost compulsory when metering packets on high-speed interfaces. This introduces some inaccuracy in the measurements. However, those flow inaccuracies can be quantified.

After the core traffic matrix is generated, it should be introduced, along with the IGP information, into a simulation application. This capacity planning approach graphically visualizes the current bottlenecks, simulates the future traffic growth and consequences, and applies what-if scenarios to verify if the SLAs such as delay and packet loss are still respected in case of link or router failure.

Part II: Implementations on the Cisco Devices