Mesh-of-Torus: a new topology for server-centric data center networks

Xie, Peibo; Gu, Huaxi; Wang, Kun; Yu, Xiaoshan; Ma, Shangqi

doi:10.1007/s11227-018-2610-4

Cited by 11 publications

(2 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Another approach is to design low-diameter networks using hierarchical composition, by combining different topologies on different levels of hierarchy. Examples of such topologies include, among others, mesh-of-trees [49], HyperX [50], [51], Petersen-based networks [52], [53], Megafly [54], mesh-of-torus [55], Slingshot [56], dragonfly-of-torus [57].…”

Section: Related Workmentioning

confidence: 99%

Realizing Optimal All-to-All Personalized Communication Using Butterfly-Based Networks

Izzi

Massini

2023

IEEE Access

View full text Add to dashboard Cite

The communication network is a key component in parallel and distributed systems, highperformance computing (HPC) systems and Data Centers, as information exchange has a key role in communication-intensive applications. The all-to-all personalized exchange is a relevant communication pattern and has been extensively studied for many network topologies. When considering a Multistage Interconnection Network (MIN) with log N stages, the realization of all-to-all communication is obtained through a sequence of admissible permutations forming a Latin square. If a particular Latin square is desired, double MINs consisting of 2 log N -1 stages are required. In this work, we consider the Butterfly-Butterfly network to realize the Latin square consisting of the identity permutation and its N − 1 rotations, RLS, and we show how to obtain the sequence of rotations forming such a Latin square, without storing it. We propose a method to realize the N permutations in pipeline fashion taking optimal O(N ) time, and we analyze the energy consumption according to the order of realization of the permutations, both on Butterfly-Butterfly network and on a single Butterfly used twice. We also show how to generate a Latin Square starting from any realizable permutation exploiting the switch configurations of the RLS.

show abstract

Section: Related Workmentioning

confidence: 99%

Realizing Optimal All-to-All Personalized Communication Using Butterfly-Based Networks

Izzi

Massini

2023

IEEE Access

View full text Add to dashboard Cite

show abstract

“…14 presented a more flexible interconnection network: Generalized-Star Cube (GSC) by combining star graph and hypercube. Xie et al proposed the Mesh-of-Torus, which incarnates the good characteristics of mesh and torus 15,16 proposed the Hybrid-TF for improving the communication performance of HPC applications based on fat-tree and Mesh-of-Torus.…”

Section: Introductionmentioning

confidence: 99%

Hybrid interconnection networks for reducing hardware cost and improving path diversity based on fat‐trees and hypercubes

Wang

2023

Concurrency and Computation

View full text Add to dashboard Cite

Fat‐tree topologies are widely used in interconnect network designs for parallel supercomputers. In the classic fat‐tree, compute nodes are connected to leaf stage switches by links. Given a large number of compute nodes, many switches and links are required, resulting in high hardware costs. To solve this problem, this paper proposes two hybrid topologies, k$$ k $$‐Cube k$$ k $$‐Ary n$$ n $$‐Tree (CAT) and Mirrored k$$ k $$‐Cube k$$ k $$‐Ary n$$ n $$‐Tree (MiCAT), based on fat‐tree and hypercube. Instead of connecting k$$ k $$ compute nodes directly to a leaf switch, we connect a k$$ k $$‐cube to the switch, and each switch in the k$$ k $$‐cube part connects k$$ k $$ compute nodes. That is, this k$$ k $$‐cube consists of 2kprefix−1$$ {2}^k-1 $$ switches and kfalse(2kprefix−1false)$$ k\left({2}^k-1\right) $$ compute nodes. We give the shortest path routing algorithms and evaluate the path diversity, cost, performance, and average packet latency of CAT and MiCAT. The results show that CAT and MiCAT can save up to 87%$$ 87\% $$ switches and 80%$$ 80\% $$ links in a large‐scale parallel system, k=n=8$$ k=n=8 $$ for example, compared to fat‐trees, and meanwhile, both CAT and MiCAT have higher path diversities than fat‐trees.

show abstract