ABSTRACT
Dragonflies are one of the most promising topologies for the Exascale effort for their scalability and cost. Dragonflies achieve very high throughput under uniform traffic, but have a pathological behavior under other regular traffic patterns, some of them very common in HPC applications. A recent study showed that randomization of task placement can make pathological, regular (multi-dimensional stencil) traffic patterns behave similar to uniform traffic.
In this work we provide a theoretical model that is able to predict the expected performance of a generic dragonfly network under uniform traffic and characterize performance-optimal dragonflies. We then analyze whether this model can be extended to other patterns by means of benchmarking the performance of multiple such patterns under both contiguous and randomized task placement. We conclude that, although in comparison with contiguous task placement, randomization does lead to a significant improvement in performance for pathological communication patterns, this performance is not on par with that of uniform traffic, but rather half of it.
- A. Bhatele, N. Jain, W. D. Gropp, and L. V. Kale. Avoiding hot-spots on two-level direct networks. In Proceedings of 2011 International Conference for High Performance Computing, Networking, Storage and Analysis, SC '11, pages 76:1--76:11, New York, NY, USA, 2011. ACM. Google ScholarDigital Library
- W. Dally and B. Towles. Principles and Practices of Interconnection Networks. Morgan Kaufmann Publishers Inc., San Francisco, CA, USA, 2003. Google ScholarDigital Library
- M. García, E. Vallejo, R. Beivide, M. Odriozola, C. Camarero, M. Valero, G. Rodríguez, J. Labarta, and C. Minkenberg. On-the-fly adaptive routing in high-radix hierarchical networks. In The 41st International Conference on Parallel Processing (ICPP), 09 2012. Google ScholarDigital Library
- J. Kim, W. J. Dally, S. Scott, and D. Abts. Technology-driven, highly-scalable dragonfly topology. In Proceedings of the 35th Annual International Symposium on Computer Architecture, ISCA '08, pages 77--88, Washington, DC, USA, 2008. IEEE Computer Society. Google ScholarDigital Library
- L. G. Valiant. A scheme for fast parallel communication. SIAM J. Comput., 11(2):350--361, 1982.Google ScholarDigital Library
Index Terms
- Randomizing task placement does not randomize traffic (enough)
Recommendations
Randomizing task placement and route selection do not randomize traffic (enough)
Dragonflies are one of the most promising topologies for the Exascale effort for their scalability and cost. Dragonflies achieve very high throughput under uniform traffic, but have a pathological behavior under other regular traffic patterns, some of ...
Performance implications of remote-only load balancing under adversarial traffic in Dragonflies
INA-OCMC '14: Proceedings of the 8th International Workshop on Interconnection Network Architecture: On-Chip, Multi-ChipDragonfly topologies are recent network designs that are considered one of the most promising interconnect options for Exascale systems. They offer a low diameter and low network cost, but do so at the expense of path diversity, which makes them ...
Quality of service using traffic engineering over MPLS: an analysis
LCN '00: Proceedings of the 25th Annual IEEE Conference on Local Computer NetworksWe compare the service received by TCP and UDP flows when they share either a link or a Multiprotocol Label Switching (MPLS) traffic trunk. Since MPLS traffic trunks allow non-shortest path links also to be used, the total network throughput goes up ...
Comments