Presentation 2003/6/13
Wormhole方式を基にしたバックトラック可能な耐故障スイッチング(ディペンダブルソフトウェアとネットワーク)
Manabu SUEISHI, Masato KITAKAMI, Hideo ITO,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Parallel computers are now popularly applied to applications where many calculations are required. In a NORA parallel computer, many processors are connected to each other by communication links and calculation results are obtained by communications between processors. It is said that the NORA computer is suitable for massive parallel computers. Since parallel computers include many processors, the failure rate of the parallel computers are higher than that of single-processor computers. From this, many fault-tolerant switching methods, which enable communications between fault-free processors even if faulty processors are included in the parallel computers, have been proposed. The existing methods have problems, however, such as low communication throughput, low fault-tolerant capability, and large hardware overhead. This paper proposes fault-tolerant switching by improving Wormhole switching, which is raid to achieve high throughput. The proposed switching inserts dummy flits after the header flit. By overwriting the header flit to the dummy flit, backtrack is implemented without hardware overhead. Computer simulations say that in a 16 by 16 2D torus, for example, the throughput of proposed method is almost equal to that of existing methods which require large hardware overhead if the number of the faulty nodes is less then 40.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) NORA parallel computer / switching / Wormhole switching / fault-tolerance / Dummy Flit / Backtrack
Paper # DC2003-6
Date of Issue

Conference Information
Committee DC
Conference Date 2003/6/13(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Dependable Computing (DC)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English)
Sub Title (in English)
Keyword(1) NORA parallel computer
Keyword(2) switching
Keyword(3) Wormhole switching
Keyword(4) fault-tolerance
Keyword(5) Dummy Flit
Keyword(6) Backtrack
1st Author's Name Manabu SUEISHI
1st Author's Affiliation Graduate School of Science and Technology, Chiba University()
2nd Author's Name Masato KITAKAMI
2nd Author's Affiliation Faculty of Engineering, Chiba University
3rd Author's Name Hideo ITO
3rd Author's Affiliation Faculty of Engineering, Chiba University
Date 2003/6/13
Paper # DC2003-6
Volume (vol) vol.103
Number (no) 134
Page pp.pp.-
#Pages 8
Date of Issue