Abstract
Random network topologies have been proposed to create low-diameter, low-latency interconnection networks in large-scale computing systems. However, these topologies are difficult to deploy in practice, especially when re-designing existing systems, because they lead to increased total cable length and cable packaging complexity. In this work we propose a new method for creating random topologies without increasing cable length: randomly swap link endpoints in a non-random topology that is already deployed across several cabinets in a machine room. We quantitatively evaluate topologies created in this manner using both graph analysis and cycle-accurate network simulation, including comparisons with non-random topologies and previously-proposed random topologies.
| Original language | English |
|---|---|
| Article number | 6860278 |
| Pages (from-to) | 2051-2060 |
| Number of pages | 10 |
| Journal | IEEE Transactions on Parallel and Distributed Systems |
| Volume | 26 |
| Issue number | 7 |
| DOIs | |
| Publication status | Published - 2015 Jul 1 |
Fingerprint
Keywords
- cabinet layout
- high-performance computing
- interconnection networks
- Network topologies
ASJC Scopus subject areas
- Hardware and Architecture
- Signal Processing
- Computational Theory and Mathematics
Cite this
Swap-And-Randomize : A Method for Building Low-Latency HPC Interconnects. / Fujiwara, Ikki; Koibuchi, Michihiro; Matsutani, Hiroki; Casanova, Henri.
In: IEEE Transactions on Parallel and Distributed Systems, Vol. 26, No. 7, 6860278, 01.07.2015, p. 2051-2060.Research output: Contribution to journal › Article
}
TY - JOUR
T1 - Swap-And-Randomize
T2 - A Method for Building Low-Latency HPC Interconnects
AU - Fujiwara, Ikki
AU - Koibuchi, Michihiro
AU - Matsutani, Hiroki
AU - Casanova, Henri
PY - 2015/7/1
Y1 - 2015/7/1
N2 - Random network topologies have been proposed to create low-diameter, low-latency interconnection networks in large-scale computing systems. However, these topologies are difficult to deploy in practice, especially when re-designing existing systems, because they lead to increased total cable length and cable packaging complexity. In this work we propose a new method for creating random topologies without increasing cable length: randomly swap link endpoints in a non-random topology that is already deployed across several cabinets in a machine room. We quantitatively evaluate topologies created in this manner using both graph analysis and cycle-accurate network simulation, including comparisons with non-random topologies and previously-proposed random topologies.
AB - Random network topologies have been proposed to create low-diameter, low-latency interconnection networks in large-scale computing systems. However, these topologies are difficult to deploy in practice, especially when re-designing existing systems, because they lead to increased total cable length and cable packaging complexity. In this work we propose a new method for creating random topologies without increasing cable length: randomly swap link endpoints in a non-random topology that is already deployed across several cabinets in a machine room. We quantitatively evaluate topologies created in this manner using both graph analysis and cycle-accurate network simulation, including comparisons with non-random topologies and previously-proposed random topologies.
KW - cabinet layout
KW - high-performance computing
KW - interconnection networks
KW - Network topologies
UR - http://www.scopus.com/inward/record.url?scp=84933575563&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84933575563&partnerID=8YFLogxK
U2 - 10.1109/TPDS.2014.2340863
DO - 10.1109/TPDS.2014.2340863
M3 - Article
AN - SCOPUS:84933575563
VL - 26
SP - 2051
EP - 2060
JO - IEEE Transactions on Parallel and Distributed Systems
JF - IEEE Transactions on Parallel and Distributed Systems
SN - 1045-9219
IS - 7
M1 - 6860278
ER -