Search for: All records

Valiant routing, the use of a random intermediate node to distribute network traffic, has been proposed for a number of recent HPC network topologies. It is also commonly used as a bulding block for adaptive routing algorithms, which use shortest path routes when possible, but revert to Valiant routing when necessary to avoid hot spots. We show that the version of Valiant routing proposed for the Slim fly topology can cause messages to follow loops, using an edge in both directions before returning to edges of the original shortest path. Removing these loops in the UGAL-L adaptive routing algorithm is shown to provide slight improvements in average latency and also allow the network to carry up to 12% more traffic before saturation. 

Dragonfly networks have been proposed to exploit high-radix routers and optical links for high performance computing (HPC) systems. Such networks divide the switches into groups, with a local link between each pair of switches in a group and a global link between each group. Which specific switch serves as the endpoint of each global link is determined by the network’s global link arrangement. We propose two new global link arrangements, each designed using intuition of how to optimize bisection bandwidth when global links have high bandwidth relative to local links. Despite this, the new arrangements generally outperform previously-known arrangements for all bandwidth relationships. 

High-performance computing systems are shifting away from traditional interconnect topologies to exploit new technologies and to reduce interconnect power consumption. The Dragonfly topology is one promising candidate for new systems, with several variations already in production. It is hierarchical, with local links forming groups and global links joining the groups. At each level, the interconnect is a clique, with a link between each pair of switches in a group and a link between each pair of groups. This paper shows that the intergroup links can be made in meaningfully different ways. We evaluate three previously- proposed approaches for link organization (called global link arrangements) in two ways. First, we use bisection bandwidth, an important and commonly-used measure of the potential for communication bottlenecks. We show that the global link arrangements often give bisection bandwidths differing by 10s of percent, with the specific separation varying based on the relative bandwidths of local and global links. For the link band- widths used in a current Dragonfly implementation, it is 33%. Second, we show that the choice of global link arrangement can greatly impact the regularity of task mappings for nearest neighbor stencil communication patterns, an important pattern in scientific applications.