|
|
|
|
|
|
|
We will describe the CPN routing algorithm which uses smart and dumb packets, and ACK packets, to sense the network state, discover paths, and bring back path information to source nodes, based on QoS driven reinforcement learning. This algorithm is the object of a recent US patent that we have been awarded, and the algorithm has been implemented in several test-beds at Imperial College and at UCF (Orlando). The test-beds use Linux based PCs as routers, and the routers themselves can be connected via Internet clouds or with point-to-point Ethernet. Our experiments show that CPN can "discover" shortest paths, but that the paths which are based on composite criteria including delay and loss can offer better QoS. We also examine issues of overhead and scalability, compatibility with IP, self-healing, and the use of QoS criteria which match the needs of Voice traffic. A power aware wireless version will also be described.
|
|
|
|
|
|
|
|