Improving Communication Load Balance Using Design Generators
Improving Communication Load Balance Using Design Generators
With the trend to massively parallel computers, the importance of communication in relation to computation constantly grows. Therefore, efficient handling of the necessary communication operations is essential. Here, the focus is on randomized routing. In agriculture, combinatorial designs have been...
Saved in:
Personal Name(s): | Burg, Heribert C. (Corresponding Author) |
---|---|
Contributing Institute: |
Jülich Supercomputing Center; JSC Zentralinstitut für Angewandte Mathematik; ZAM |
Published in: | 1996 |
Imprint: |
Jülich
Zentralinstitut für Angewandte Mathematik
1996
|
Physical Description: |
10 p. |
Document Type: |
Report |
Research Program: |
ohne Topic |
Publikationsportal JuSER |
With the trend to massively parallel computers, the importance of communication in relation to computation constantly grows. Therefore, efficient handling of the necessary communication operations is essential. Here, the focus is on randomized routing. In agriculture, combinatorial designs have been invented to guarantee fair experiment results by omitting biases caused by varying soil fertilities. In other words, the construction principle of the experimental designs guaranteed a balanced spread of the different crops over the test field. Exactly these originary properties of designs will be used here to guarantee an optimal load balancing of communication traffic in an interconnection network. Difference sets are a means to generate an important kind of combinatorial designs. Simulation experiments prove the superiority of the difference set based routing schemes over standard routing schemes. |