Trace-driven co-simulation of high-performance computing systems using OMNeT++
Proceedings of the 2nd International Conference on Simulation Tools and Techniques
HyperX: topology, routing, and packaging of efficient large-scale networks
Proceedings of the Conference on High Performance Computing Networking, Storage and Analysis
LogGOPSim: simulating large-scale applications in the LogGOPS model
Proceedings of the 19th ACM International Symposium on High Performance Distributed Computing
Hedera: dynamic flow scheduling for data center networks
NSDI'10 Proceedings of the 7th USENIX conference on Networked systems design and implementation
Improving datacenter performance and robustness with multipath TCP
Proceedings of the ACM SIGCOMM 2011 conference
Bandwidth-optimal all-to-all exchanges in fat tree networks
Proceedings of the 27th international ACM conference on International conference on supercomputing
Channel reservation protocol for over-subscribed channels and destinations
SC '13 Proceedings of the International Conference on High Performance Computing, Networking, Storage and Analysis
Scalable, optimal flow routing in datacenters via local link balancing
Proceedings of the ninth ACM conference on Emerging networking experiments and technologies
Hi-index | 0.00 |
Today's scalable high-performance applications heavily depend on the bandwidth characteristics of their communication patterns. Contemporary multi-stage interconnection networks suffer from network contention which might decrease application performance. Our experiments show that the effective bisection bandwidth of a non-blocking 512-node Clos network is as low as 38% if the network is routed statically.In this paper, we propose and analyze different adaptive routing schemes for those networks. We chose Myrinet/MX to implement our proposed routing schemes. Our best adaptive routing scheme is able to increase the effective bisection bandwidth to 77% for 512 nodes and 100% for smaller node counts. Thus, we show that our proposed adaptive routing schemes are able to improve network throughput significantly.