GNS3 simulation tests of BGP delays in re‐routing for IP Anycast

  • Author(s) / Creator(s)
  • BGP and IP Anycast are two major technologies used by Internet Service Providers to support their various services such as, Hosting, FTP, e‐mail, Data Storage, Printer Server. BGP is mainly used to support routing decisions in the core of Internet and permits forwarding packets among autonomous
    systems. On the other hand, IP Anycast is used to allow different providers to have mirror sites and to allow for load balancing and backup/restore points in the network. This project focuses on analyzing the convergence time when using BGP as a unique routing protocol and, in combination with IP Anycast, as a tool to re‐route traffic to another point in the network. For example, when an outage occurs and one of the IP Anycast servers is down, the network has to find a new route to the destination. Using BGP’s process, the network must wait until the path is withdrawn, and then a new path is established to the server. It is possible that this process could take more than 180 seconds in default conditions to be able to transmit data again to the
    same service in the new location. The main finding of this project was that convergence time is close to the hold‐time setting for BGP. Moreover this convergence time is higher than the convergence time for IP Anycast re‐routing when using OSPF as the routing protocol. Also when the default BGP timer values were changed, the same tendency in convergence time was confirmed. Thus, I conclude that anycast convergence time is related to the value of the BGP hold‐timer. In addition, I found that some sample convergence times are higher than the value of the hold‐timer, and this situation occurs more often when the path to the destination is longer.

  • Date created
    2011-02-01
  • Subjects / Keywords
  • Type of Item
    Report
  • DOI
    https://doi.org/10.7939/r3-1et8-zn25
  • License
    Attribution-NonCommercial 4.0 International