Search

In this paper, we propose an approximate load balance approach based on the Identifler(ID)/Locator split routing architecture. In our approach, we propose a multipath method and a schedule algorithm. Both of them run on each router for scheduling each packet that is received from the connected host onto the proper path to the destination host. Our approach allows each router using the ID/Locator splitting to obtain the multiple paths between each pair of hosts. The merit of our approach is that the path is selected by each router individually to achieve load balance other than a central controller in the ID/Locator split routing architecture. The time complexity of the schedule algorithm is much lower than the linear programming (LP) and nonlinear programming (NLP). Simulation results show that, compared to the existing single path routing based on the ID/Locator split routing architecture, the mean of normalized link utilization (NLU) increases by more than 75% relatively on average of all simulated scenarios.

In this paper, we address the issue of reducing intra-domain network configuration. To achieve this, we design an OSPF based routing protocol, named Reducing Complexity of Network Configuration (RCNC), which works with DHT based Identifier(ID)/Locator split routing mechanism RCNC router automatically generates Router ID for itself, and creates Router ID based routing table by exchanging link-state advertisements (LSA) with others, without configuring IP and subnet mask on the interface that links with other routers. Meanwhile, the ID/Locator split based routing mechanism ensures the RCNC routers backward compatible with existing hosts. Evaluation results show that in 10 actual ISPs, RCNC reduces an average of 13.3 items of intra-domain configuration per router when the total number of configuration reduction is minimum, and 6.8 items when the total number is maximum. Meanwhile, comparing with the existing routing protocol, in the 10 ISPs, RCNC reduces intra-domain routing entries by 67.5% on average when the number of entries is minimum and 27.5% on average when the number is maximum. Through the NS-2 simulation, we can observe that the control overhead of RCNC only increases by 1.9% than that of OSPF.