Twisted and Folded Clos-Network Design Model With Two-Step Blocking Probability Guarantee
Haruto Taka, Takeru Inoue, Eiji Oki
Abstract
Haruto Taka, Takeru Inoue, Eiji Oki
Abstract
A data center network often employs a Clos-network structure due to scalability nature. A previous design model analyzed blocking probability to guarantee the quality of Clos network. However, the conditions to guarantee the blocking probability in these models are stringent, which shrinks the switching capacity. This letter proposes a model that relaxes network design constraints to maximize switching capacity while guaranteeing blocking probability in a two-step manner, where the proposed model divides intermediate switches into two groups. Numerical results show that the proposed model achieves a larger switching capacity than a conventional one-step model.
OpenAlex reports 14 citations for this work. Citation counts describe recorded attention and do not establish research quality.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
A data center network often employs a Clos-network structure due to scalability nature. A previous design model analyzed blocking probability to guarantee the quality of Clos network. However, the conditions to guarantee the blocking probability in these models are stringent, which shrinks the switching capacity. This letter proposes a model that relaxes network design constraints to maximize switching capacity while guaranteeing blocking probability in a two-step manner, where the proposed model divides intermediate switches into two groups. Numerical results show that the proposed model achieves a larger switching capacity than a conventional one-step model.
Key concepts: Clos network, Blocking (statistics), Computer science, Computer network