This chapter discusses how to test the super cluster (see definition in Appendix). Initially, you will shut down the entire supercomputer by sequentially using the shutdown -h now command at each node, and then reenergize the super cluster to reinitialize the machine as it progresses through its start up sequence. After the nodes are up-and-running, you will then engage multiple nodes using the -H command to run/test the super cluster by eviscerating the time need to solve the π function. Finally, you will create convenience bash files to facilitate ease of use of your supercomputer.
In this chapter, you will learn:
- How to use the
shutdown -h nowcommand to shut down your Pis - How to engage multiple cores/nodes, using the
-Hcommand, to solve the MPIπfunction in record time - How to create
convenience bashfiles for enhancing the user experience, while operating the supercomputer
Shut down the entire cluster by successively ssh-ing into each node, and enter sudo shutdown -h now (we will discuss automating this procedure later). Reenergize the cluster by turning off the power strip, and then on again, or unplug the rapid charger from the wall socket, and then reinsert the plug into the socket. Wait about 30 seconds or so for the Pis to reboot. Then ssh into the master node, and switch to the alpha user, then change the directory to the gamma folder.
Execute the call-procs program to see if all the nodes are working appropriately. For example, if you have an eight or 16-node machine, type in, individually; mpiexec -H Mst0 call-procs, mpiexec -H Slv1 call-procs, mpiexec -H Slv2 call-procs, mpiexec -H Slv3 call-procs...mpiexec -H Slv7 call-procs or Slv15 call-procs. This procedure will initialize the nodes. Then command all the nodes using the command; mpiexec -H Slv1,Slv2,Slv3,Slv4,Slv5,Slv6,Slv7,Slv8,Slv9,Slv10,Slv11,Slv12,Slv13,Slv14,Slv15 call-procs. If all the nodes respond appropriately, then you are good to go. Have fun!
Fellow geeks, this is now the decisive moment. Rerun the command for testing the master, and slave Pi; that is, enter the command time mpiexec -H Mst0,Mst0,Mst0,Mst0,Slv1,Slv1,Slv1,Slv1 MPI_08_b. The runtime should be similar to your earlier runtime of approximately 14m19.05s, using 300,000 iterations in the restrained mode. After the run ends, run the nodes as shown on the following pages. Engaging the entire stack of eight Pi2s, should produce a run time of approximately 6m58.102s. If you observe this result, you have then achieved your goal of building a miniature supercomputer so be proud of your accomplishment. You have my permission to pop the Champagne cork now. You have just built your first eight or 16-node Raspberry Pi2 or Pi3 supercomputer. So, go ahead now, and wield your powerful -H scepter to demolish any or all mathematical enemies that stand in your path. Start with the runs depicted on the following pages.