Adaptive load balancing approach for Cluster Bidirectional Diagonal Mesh Topology

In multicore embedded system power and performance of system are major concern. The design of network architecture and routing algorithm mainly determines the performance of the Network on Chip. In Q routing each router stores large tables due to this there is large area overhead. To reduce this overhead clustering of Bidirectional Diagonal Mesh network is performed. This motivates to design Cluster Bidirectional Diagonal Mesh architecture and a routing algorithm known as Cluster Algorithm using Modified Q routing and Diagonal routing (CAMQD) to support this architecture. So a new architecture Cluster Bidirectional Diagonal Mesh along with the routing algorithm for this architecture is proposed in this paper. Scalability can be achieved with clustering. CAMQD is congestion aware routing algorithm with low latency and power requirement. It is fully adaptive and at the same time, it is fault tolerant and can handle deadlock situation. CAMQD is a quasi-minimal algorithm. The performance is analyzed and compared with mesh topology using Q routing in terms of latency, throughput and power with different packet injection ratio using Constant Bitrate Traffic (CBR). This topology offers parallelism and scalability to utilize full potential of NoC.