I/O Node Placement for Performance and Reliability in Torus Networks

B. Azeez, H. Kim, Y. Jin, and E.J. Kim (USA)


Resource Allocation, Torus Interconnection Networks, Fault Tolerance


When a cluster system interconnects processor and I/O nodes through a network, an optimal placement of I/O nodes is critical to improve the overall system perfor mance by reducing its communication latency. In this pa per, we propose an efficient and scalable I/O node place ment scheme, called a relaxed quasi-perfect, for torus based interconnection networks using Lee distance error correcting code. It provides a more general placement than the previous quasi-perfect placement [1]. We also suggest a fault-tolerant scheme using our I/O placement model for a guaranteed performance. Simulation results show that our scheme provides 53% speed-up over the quasi-perfect. Also the fault tolerant scheme provides a graceful slow down especially until the number of faulty I/O nodes be comes less than half of the initial I/O nodes.

Important Links:

Go Back