Narrow Results

As the key component of wireless network-on-chip (WiNoC), wireless router is required to handle a large number of data packets which could cause network congestion. The congestion not only reduces the network performance, but also results in additional power consumption. In this paper, a low-power WiNoC with congestion-aware wireless node is proposed. Firstly, flit counter unit and address resolution unit are added in wireless interface (WI), which can sense the congestion information and destination address information of each wireless node dynamically. Secondly, through the proposed congestion judgment algorithm, the congestion judgment unit in global network can judge the priority of each wireless node pair and set the highest-priority wireless node pair to use the channel resources preferentially. The mechanism can effectively alleviate the congestion of wireless nodes and ensure the stability of system performance. In addition, a sleep mechanism is utilized to switch off WI which fails in the wireless channel competition. Experimental results show that the proposed scheme can effectively improve the performance in terms of packets’ transmission latency and network throughput.

Wireless on-chip networks suffer from serious congestion problems due to the expansion of their network size and the joining of wireless nodes, and token passing is very inefficient in the case of low network injection. Aiming at the congestion problem and the inefficiency of token passing in wireless on-chip networks, this paper designs a three-hop congestion-aware routing mechanism in WiNoC. In the scheme of this paper, first, the congestion information is written into Head Flit for propagation through the information piggybacking method, in which the information piggybacking method does not introduce congestion information propagation overhead, second, the port with the smallest congestion value is selected for transmission through the three-hop congestion-aware routing algorithm to reduce the congestion within the subnetwork, and then the wireless node congestion is mitigated through the new wired and wireless packet division method, and finally, the power of wireless nodes is reduced through the dynamic MAC mechanism to reduce the power consumption of wireless nodes. The experimental results show that this paper’s scheme can reduce the intra-subnet congestion, balance the inter-subnet load, reduce the power consumption of wireless nodes, and reduce the latency by 44.6% compared with the basic wireless on-chip network using adaptive routing algorithm, and reduce the latency by 21.8% compared with the wireless on-chip network using global congestion-aware routing algorithm; in different traffic patterns, this paper achieves the highest saturation throughput in different traffic patterns, and the additional hardware overhead of the router increases by only 0.04%.

The issue of routing is vitally important in Wireless Mesh Networks. But currently, most routing protocols such as OLSR and AODV cannot make the most of multiple paths between the source site and destination site because of the complexity and the cost. Software Defined Networking(SDN) structure promises to obtain the network configuration effectively, and with a centralized controller, it can deploy fine-grained routing algorithms to make full use of the network resources, while ensuring that the control overhead is acceptable. In this paper, we propose a new approach of SDN-based routing algorithm, or SDNR. We introduce link saturation to SDN, with which the SDN controller can figure out the congested path and reroute the following traffic to another non-congested path, which is the real-time optimal one, to ensure the network throughput. We compare SDNR to classic routing protocols and demonstrate its superiority.