Jan. 28, 2013 to Jan. 30, 2013
Menghan Li , Dept. of Comput. Sci. & Technol., Tsinghua Univ., Beijing, China
Xiaoping Zhang , Dept. of Comput. Sci. & Technol., Tsinghua Univ., Beijing, China
Yupeng Tian , Dept. of Comput. Sci. & Technol., Tsinghua Univ., Beijing, China
Haixiang Zhang , Dept. of Comput. Sci. & Technol., Tsinghua Univ., Beijing, China
The centralized switching fabric is increasingly becoming the performance bottleneck of the routers due to the constraints of the port size and the complexity of scheduling algorithm. In order to address scalability problem, interconnection networks have been widely used to build distributed switching fabrics. P2i is an asymptotically minimal node-degree interconnection network which has an ideal throughput of at least 2 under uniform traffic pattern for arbitrary node numbers. We propose B-P2i- architecture, which enhances the path diversity and shortens the network diameter of P2i. In order to balance the transmission efficiency under light load and the maximum throughput under heavy load, we present PB routing algorithm for B-P2i- architecture which is constructed by IQ routers. Our simulations using the modified BookSim simulator indicate that B-P2i- architecture using PB routing algorithm has good performance under adversary traffic patterns. The simulations also indicate that our architecture has an ideal scalability for constructing scalable routers with large capacity. To the best of our knowledge, this is the first time to utilize the concept of potential field to design adaptive routing algorithm for scalable router which is constructed by IQ routers with DIN topology.
Topology, Routing, Throughput, Algorithm design and analysis, Network topology, Computer architecture, Force,
Menghan Li, Xiaoping Zhang, Yupeng Tian, Haixiang Zhang, "Potential based routing in B-P2i− architecture", ICOIN, 2013, The International Conference on Information Networking 2014 (ICOIN2014), The International Conference on Information Networking 2014 (ICOIN2014) 2013, pp. 71-76, doi:10.1109/ICOIN.2013.6496354