Technology-Driven, Highly-Scalable Dragonfly Topology

By John Kim, William (Bill) J. Dally, S. Scott, D. Abts. International Symposium on Computer Architecture, 2008

850 citations recorded by the source corpus.

Subfields: Computer Networks and Communications, Electrical and Electronic Engineering

View Paper: https://www.semanticscholar.org/paper/b81350c2601441dcd526e656a590ff354dedb220

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Networks and Communications

1. Torsten Hoefler
2. Maciej Besta
3. Brighten Godfrey
4. C. Hong
5. Fan Yang

In Electrical and Electronic Engineering

1. Torsten Hoefler
2. Maciej Besta
3. Brighten Godfrey
4. Han Hu
5. C. Hong

Citations in the corpus

• Jellyfish: Networking Data Centers Randomly (2011, background): “We limit our interest in cable length to whether or not a cable is short enough, i.e., 10 meters in length [17, 26], for use of an electrical cable.”
• Slim Fly: A Cost Effective Low-Diameter Network Topology (2014, methodology): “Each group is fully connected (p(p−1)2 electric channels) and there are p fiber links between every two groups in the same row or column of racks. e) Dragonfly and Random Networks: We use the balanced DF [28] (a = 2p = 2h). a is the number of routers in a group and h is the number of fiber cables co”
• Tutel: Adaptive Mixture-of-Experts at Scale (2022, methodology): “For large-scale network topologies like dragonfly (Kim et al., 2008), 2DH All-to-All could be further adapted to 3D by splitting inter-node to intra-group and inter-group All-to-All according to the network hierarchy.”