Fri, 01/25/2008 - 12:12 by Damien Leroy • Categories:


Once multihomed, a site usually wants to engineer its interdomain traffic. In the IDIPS ("ISP-Driven Informed Path Selection") approach, hosts delegate the choice of the best source and destinations addresses to use to contact a destination node to a special server. By selecting these addresses, the IDIPS server roughly determines the routing. It thereby provides features like traffic engineering and fault tolerance, without transmitting any BGP advertisement and without impacting on the worldwide routing table size.

In his early version, IDIPS was proposed for multihomed sites, but it can also be used to select the best server among a list of duplicated servers accordingly to the ISP considerations of cost, performances or policies.

Fig. 1 shows a global overview of an IDIPS deployment. IDIPS can receive information from active or passive measurements, routing protocols or even network policies.

Figure 1 - IDIPS global deployment overview

Fig. 2 shows the IDIPS as implemented in XORP

Figure 2 - IDIPS Architecture

IDIPS is decomposed in three modules. The Querying module is in charge of interacting with the clients and ranking the path. The Measurement module is in charge of measuring the network either actively (e.g., ping) or passively (e.g., NetFlow). The measurement module can also read information from the routing protocols. Finally, the Prediction module uses the past measurements to infer the future behavior of the network. The querying module uses the predictions made by the prediction module to determine the rank the paths requested by the clients. Both the prediction and the measurements modules can run several measurements and prediction in parallel. For example, the prediction module can measure the delay to important destinations, the available bandwidth to some backup servers and the loss rate to the VoIP gateways.

Project material

Related works