Performance and fault tolerance of the ideal system network via additional parallelity
Автор: Viktor S. Podlazov
Журнал: Программные системы: теория и приложения @programmnye-sistemy
Рубрика: Программное и аппаратное обеспечение распределенных и суперкомпьютерных систем
Статья в выпуске: 3 (66) т.16, 2025 года.
Бесплатный доступ
The article considers methods for increasing the performance of system networks by increasing their parallelism in two ways: using their copies and parallel channels in each copy. The article considers methods for constructing non-blocking system networks with direct channels of arbitrary size. The article considers methods for constructing non-blocking networks with a given number of parallel channels as networks with the structure of quasi-complete graphs. The article considers a method for constructing system networks of any size with a given number of parallel channels by using the corresponding number of their copies.
Complete switch, direct channels, multiplexers and demultiplexers, multi-stage switch, conflictfree routing, non-blocking switch, static self-routing, quasi-complete graph with a given number of parallel channels, isomorphism of a quasi-complete graph and a symmetric block-design
Короткий адрес: https://sciup.org/143184622
IDR: 143184622 | DOI: 10.25209/2079-3316-2025-16-3-41-68