2 Replies Latest reply: May 12, 2010 10:33 AM by sen RSS

    BGP partial or full or default




      I have a question about multihome BGP. I know that we can get just defualt route, partial or full routing table from the ISP in to our enterprise EDge but trying to undesratnd what is advantage of one over other in a load sharing enviroment? Just getitng the default route wouldn't do the trick ? what scenarios i need to get the full routing table and the advantage the full table?


      Appreciate your inputs/clarrifications

        • 1. Re: BGP partial or full or default
          Scott Morris - CCDE/4xCCIE/2xJNCIE

          How much memory do you have?   How much do you want to truly look at  when doing "show ip bgp" or "show ip route"?  How many paths do you have  and how much variance to you really anticipate having?  What CPU are we  talking about?


          Unless you are working on some serious  traffic engineering/balancing, or have a need for separation that much,  most people with multihoming to different ISPs are perfectly cool with  partial routing.  You'll get the "customer routes" for that particular  ISP and a default.


          MOST routing would likely be the  default route and would balance out.


          Full routes is  overwhelming for most people and little to no gain is seen.  But  certainly analyze what you are doing and what you want to see!



          • 2. Re: BGP partial or full or default
            Tammy B

            Have you looked at PfR?  Just throwing it out there as it's an older, yet emerging technology that could do your load balancing based on several network related factors.


            OER was developed to identify and control network performance issues that traditional IP routing cannot address. In traditional IP routing, each peer device communicates its view of reachability to a prefix destination with some concept of a cost related to reaching the metric. The best path route to a prefix destination is usually determined using the least cost metric, and this route is entered into the routing information base (RIB) for the device. As a result, any route introduced into the RIB is treated as the best path to control traffic destined for the prefix destination. The cost metric is configured to reflect a statically engineered view of the network, for example, the cost metric is a reflection of either a user preference for a path or a preference for a higher bandwidth interface (inferred from the type of interface). The cost metric does not reflect the state of the network or the state of the performance of traffic traveling on that network at that time. Traditional IP routed networks are therefore adaptive to physical state changes in the network (for example, interfaces going down) but not to performance changes (degradation or improvement) in the network. Occasionally, degradation in traffic can be inferred from either the degradation in performance of the routing device or the loss of session connectivity, but these traffic degradation symptoms are not a direct measure of the performance of the traffic and cannot be used to influence decisions about best-path routing.

            To address performance issues for traffic within a network, OER manages traffic classes. Traffic classes are defined as subsets of the traffic on the network, and a subset may represent the traffic associated with an application, for example. The performance of each traffic class is measured and compared against configured or default metrics defined in an OER policy. OER monitors the traffic class performance and selects the best entrance or exit for the traffic class. If the subsequent traffic class performance does not conform to the policy, OER selects another entrance or exit for the traffic class.