MAPS: a localized and distributed adaptive path selection scheme

The issue of how to select a label switched path (LSP) for a flow request in multiprotocol label switching (MPLS) networks to meet QoS requirements, such as throughput and end-to-end delay, has attracted a lot of attention from researchers. Most existing QoS routing schemes, such as the widest-shortest-path (WSP) scheme, need frequent updates of global network link status since stale information can drastically degrade their performance. We propose a localized and distributed MPLS adaptive path selection (MAPS) scheme. The proposed scheme makes path selection and routing decisions only at the edge routers based on local information, such as the blocking probability observed at the edge routers. These decisions aim to minimize the network congestion and reduce blocking probability in the whole network. Through extensive simulation, the MAPS scheme is shown to be effective in significantly reducing the blocking probability without core network information, when compared to the widest-shortest-path and random path selection schemes.