Dynamical traffic analysis on scale-free networks

Evaluation of trade-offs between dynamical characteristics and architecture design on large-scale networks is demanded for engineering solutions. We study dynamical characteristics of complex networks through both packet traffic simulation and theoretical network analysis to explore highperformance NoC (network-on-chip) architectures. To achieve efficient communication with hundreds of functional modules implemented on a chip in the future, scale-free network is considered as one of the promising topologies for constructing on-chip networks. In scale-free networks, we focus on pair degree-degree correlations that indicate average degree of the nearest neighbors of a node, and investigate their topological effects on NoC performance. The network architecture constructed with the topology such as hubs mostly connect to lower-degree nodes can disperse traffic load efficiently and avoid the extreme concentration of load on hubs.