As the size of high-performance computing systems grows, the number of events requiring a network reconfiguration, as well as the complexity of each reconfiguration, is likely to increase. In large systems, the probab...
详细信息
ISBN:
(纸本)9781467365987
As the size of high-performance computing systems grows, the number of events requiring a network reconfiguration, as well as the complexity of each reconfiguration, is likely to increase. In large systems, the probability of component failure is high. At the same time, with more network components, ensuring high utilization of network resources becomes challenging. Reconfiguration in interconnection networks, like InfiniBand (IB), typically involves computation and distribution of a new set of routes in order to maintain connectivity and performance. In general, current routingalgorithms do not consider the existing routes in a network when calculating new ones. Such configuration-oblivious routing might result in substantial modifications to the existing paths, and the reconfiguration becomes costly as it potentially involves a large number of source-destination pairs. In this paper, we propose a novel routingalgorithm for IB based fat-tree topologies, SlimUpdate. SlimUpdate employs techniques to preserve existing forwarding entries in switches to ensure a minimalrouting update, without any performance penalty, and with minimal computational overhead. We present an implementation of SlimUpdate in OpenSM, and compare it with the current de facto fat-tree routingalgorithm. Our experiments and simulations show a decrease of up to 80% in the number of total path modifications when using SlimUpdate routing, while achieving similar or even better performance than the fat-tree routing in most reconfiguration scenarios.
The performance of NoCs (Networks-On-Chip) highly relies on the routingalgorithm. Despite the higher implementation complexity compared with deterministic routing, adaptive routing has several merits, such as lower l...
详细信息
ISBN:
(纸本)9781467314824;9781467314817
The performance of NoCs (Networks-On-Chip) highly relies on the routingalgorithm. Despite the higher implementation complexity compared with deterministic routing, adaptive routing has several merits, such as lower latency, higher throughput and better fault-tolerance performance. Most of the existing adaptive routingalgorithms are based on the comparison of the horizontal and vertical congestion status in the network. However the performance of adaptive routing schemes suffers from the inadequate global congestion information. To address this issue, we proposed a novel routingalgorithm with a congestion aware subnetwork to obtain more accurate non-local congestion information. This subnetwork will propagate the congestion information along the paths directly towards the destination. To find a less congested path, PARS (Path-Aware routing Scheme) uses the congestion information of paths in a straight direction towards the destination rather than local congestion information. The simulation results reveal that the new presented scheme can offer better performance under different traffic profiles with a small hardware overhead.
暂无评论