NoSwitchport.com

Optimized Edge Routing Overview

Posted in Networking by shaw38 on January 8, 2010

If you can tell me of a more understated topic on the CCIE Routing and Switching v4.0 lab blueprint than Optimized Edge Routing (OER), I’ll buy you a beer. This was quietly snuck into the blueprint in between policy-based routing and redistribution, both fairly straightforward topics. Should be no big deal right? False.

OER removes rigidness of standard IP routing where typical routing metrics are derived from physical layer measurements and in turn, dictates a generic routing policy for all traffic. OER does this by gathering higher-lever performance metrics through IP SLA and Netflow information and uses this to determine the most optimal exit point for certain destination prefixes or traffic classes. Once the ideal exit point has been decided, routing policy is dynamically updated to influence the specific traffic class.

Navigating through the configuration guide for OER can be daunting but configuration can be broken down into 5 steps:

1. Profile

  • The selection of a subset of traffic to optimize performance
  • Learns the flows passing through the router with the highest delay or throughput
  • Statically configure class of traffic to performance route

2. Measure

Once traffic has been profiled, metrics need to be generated against it. This is down through:

  • Passive monitoring – measuring performance of a traffic flow as the flow is traversing the data path
  • Active monitoring – generating/measuring synthetic traffic to emulate the traffic class being monitored
  • Both can be deployed: passive monitoring can be used to determine if the flow doesn’t conform to an oer policy and active monitoring can find the most optimized alternate path

3. Apply Policy

  • Performance metrics are compared to a set of low and high thresholds and a determination is made if the metrics are out of policy
  • Traffic class policies – defined for prefixes or for applications
  • Link policies – defined for entrance or exit links at the edge

4. Control

  • Traffic flow is modified to enhance network performance
  • Methods for modifying routing policy:
    • For traffic classes defined using a specific prefix, traditional routing information can be modified using BGP or an IGP to add/remove a route
    • For traffic classes defined by application (prefix + upper layer protocol), there are two methods:
      • Device specific: Policy-based routing
      • Network specific:
        • Overlay performance – MPLS or mGRE to reach any other device at the network edge
        • Context Enhanced Protocols – BGP/OSPF/EIGRP are enhanced to communicate upperlayer information with a prefix

5. Verify

  • Once traffic is flowing through the perferred exit point, the traffic class is verified again against the traffic policy
  • If it is determined to the traffic is still out of profile, the controls put in place are reverted and the measurement phase restarts
Tagged with: , ,