Best Practices for Determining the Traffic Matrix in IP Networks

 

• Internal Traffic Matrix

   – PoP to PoP matrix

      • Can be from core (CR) or access (AR) routers

   – Class based

• External Traffic Matrix

    – PoP to External AS

          • BGP

          • Origin-AS or Peer-AS

                 – Peer-AS sufficient for Capacity Planning and Resilience Analysis

           • Useful for analyzing the impact of external failures on the core network (capacity/resilience)

• Example Data from Tier-1 IP Backbone

   – Measured Traffic Matrix (MPLS TE based)

   – European and American subnetworks

   – 24h data

• Properties

   – Temporal Distribution

       • How does the traffic vary over time

   – Spatial Distribution

       • How is traffic distributed in the network?

   – Relative Traffic Distribution

       • “Fanout”

Traffic Collection:

• Data is collected at fixed intervals

    – E.g. every 5 or 15 minutes

• Measurement of Byte Counters

    – Need to convert to rates

    – Based on measurement interval

• Create Traffic Matrix

   – Peak Hour Matrix

      • 5 or 15 min. average at the peak hour

   – Peak Matrix

      • Calculate the peak for every demand

      • Real peak or 95-percentile

Collection Methods:

• NetFlow

     – Routers collect “flow” information

    – Export of raw or aggregated data

• DCU

    – Routers collect aggregated destination statistics

•  MPLS

    – LDP

          • Measurement of LDP counters

     – RSVP

          • Measurement of Tunnel/LSP counters

• Estimation

     – Estimate Traffic Matrix based on Link Utilizations

Reference: http://www.ltcm.net/~telkamp/papers/p149-gunnar.pdf