Details about the 24 node cluster we built are included in the following links
1. http://blog.afkham.org/2013/01/raspberry-pi-control-center.html
2. http://blog.afkham.org/2013/02/building-raspberry-pi-cluster-part-2.html
Details about the 24 node cluster we built are included in the following links
1. http://blog.afkham.org/2013/01/raspberry-pi-control-center.html
2. http://blog.afkham.org/2013/02/building-raspberry-pi-cluster-part-2.html
For comparison, there is another recent 24-node RPi cluster report at:
http://eclipsejpa.blogspot.com/2012/11/java-7-se-parallel-processing-cluster.html
It notes that the mpi pi-calculation benchmark scales very badly, running slower
with each additional node.
1 node: wall clock time = 0.017286
2 node: wall clock time = 0.020435
4 node: wall clock time = 0.037727
6 node: wall clock time = 0.043331
This is the same benchmark that the University of Southampton used,
although their published results were for only two nodes, and omitted the timings.
http://www.southampton.ac.uk/~sjc/raspberrypi/pi_supercomputer_southampton.htm
Thanks for posting coder27.
Thanks for posting coder27.