Twisted Torus Topologies for Enhanced Interconnection Networks

被引:35
作者
Camara, Jose M. [1 ]
Moreto, Miquel [2 ]
Vallejo, Enrique [3 ]
Beivide, Ramon [3 ]
Miguel-Alonso, Jose [4 ]
Martinez, Carmen [3 ]
Navaridas, Javier [4 ]
机构
[1] Univ Burgos, Dept Electromech Engn, Burgos 09006, Spain
[2] Univ Politecn Cataluna, Dept Comp Architecture, ES-08034 Barcelona, Spain
[3] Univ Cantabria, Dept Elect & Comp, E-39005 Santander, Spain
[4] Univ Basque Country, Dept Comp Architecture & Technol, Donostia San Sebastian 20018, Spain
关键词
Multiprocessor interconnection; parallel architectures; routing; supercomputers;
D O I
10.1109/TPDS.2010.30
中图分类号
TP301 [理论、方法];
学科分类号
081202 ;
摘要
Many current parallel computers are built around a torus interconnection network. Machines from Cray, HP, and IBM, among others, make use of this topology. In terms of topological advantages, square (2D) or cubic (3D) tori would be the topologies of choice. However, for different practical reasons, 2D and 3D tori with different number of nodes per dimension have been used. These mixed-radix topologies are not edge symmetric, which translates into poor performance due to an unbalanced use of network resources. In this work, we analyze twisted 2D and 3D mixed-radix tori that remove the network bottlenecks present in nontwisted ones. Such topologies recover edge symmetry, and consequently, balance the utilization of their links. The distance-related properties of twisted tori together with a full characterization of their bisection bandwidth are described in this paper. A simulation-based performance evaluation has been carried out to assess the network performance under synthetic and trace-driven workloads. The obtained results show noticeable and consistent performance gains (up to an increase of 74 percent in accepted load). In addition, we propose scalable and practicable packet routing mechanisms and wiring layouts for these interconnection systems. The complexity of the architectural proposals is similar to the one exhibited by routing and folding mechanisms in standard tori.
引用
收藏
页码:1765 / 1778
页数:14
相关论文
共 26 条
[1]   Blue Gene/L torus interconnection network [J].
Adiga, NR ;
Blumrich, MA ;
Chen, D ;
Coteus, P ;
Gara, A ;
Giampapa, ME ;
Heidelberger, P ;
Singh, S ;
Steinmacher-Burow, BD ;
Takken, T ;
Tsao, M ;
Vranas, P .
IBM JOURNAL OF RESEARCH AND DEVELOPMENT, 2005, 49 (2-3) :265-276
[2]  
ADIGA NR, 2002, P ACM IEEE C SUP SUP
[4]  
Beivide R., 1987, P 14 ANN INT S COMP, DOI [10.1145/30350.30369, DOI 10.1145/30350.30369]
[5]   ILIAC IV SYSTEM [J].
BOUKNIGHT, WJ ;
SAMEH, AH ;
SLOTNICK, DL ;
MCINTYRE, DE ;
DENENBERG, SA ;
RANDALL, JM .
PROCEEDINGS OF THE INSTITUTE OF ELECTRICAL AND ELECTRONICS ENGINEERS, 1972, 60 (04) :369-+
[6]   Packaging the blue Gene/L supercomputer [J].
Coteus, P ;
Bickford, HR ;
Cipolla, TM ;
Crumley, PG ;
Gara, A ;
Hall, SA ;
Kopcsay, GV ;
Lanzetta, AP ;
Mok, LS ;
Rand, R ;
Swetz, R ;
Takken, T ;
La Rocca, P ;
Marroquin, C ;
Germann, PR ;
Jeanson, MJ .
IBM JOURNAL OF RESEARCH AND DEVELOPMENT, 2005, 49 (2-3) :213-248
[7]  
*CRAY INC, 2008, CRAY X1E SUP
[8]  
*CRAY INC, 2008, CRAY XT4 DAT
[9]  
*CRAY INC, 2008, CRAY XT3 DAT
[10]   Performance analysis of the alpha 21364-based HP GS1280 multiprocessor [J].
Cvetanovic, Z .
30TH ANNUAL INTERNATIONAL SYMPOSIUM ON COMPUTER ARCHITECTURE, PROCEEDINGS, 2003, :218-228