Timeline

Organization of Data

You can find the detailed results and the explanation of the concepts used in the results section.

May 11, 2006

We packed the served and returned back to Sun. Although we have a lot of ideas to implement, the 60 day period finalized. It was an exciting experience!

May 10, 2006 New !

We are comparing T2000 with another parallel computer. Results will be posted soon.

May 8, 2006

We discovered the SWaP (Space, Watts and Performance) metric. Sun created SWaP to help decision makers quantify how much power, air conditioning and space a server consumes. It can be calculated using the following formula:

From this moment we will include this metric in our results.

April 28, 2006

We are running new tests on T2000. Results will be posted soon.

April 26, 2006

We plotted the results in 3D surfaces. This visualization technique allow us to see better the dynamic behavior of T2000. See results for more details.

April 21, 2006

We obtained a new set of results. For more details, see the results section.

April 18, 2006

We installed LAM/MPI successfully. The package comes with a testsuite. One of the test was not passed (attrsetget_f). MPI scales well up to 32 processes, but drops off after that. We are still working on a hybrid MPI/OpenMP tester.

April 14, 2006

The change of the parameter made a huge difference for higher numbers of threads. Performance was about the same for 1, 2, 4, and 8 threads. But from 16 threads, there was a big improvement. For more details, see the results section.

April 13, 2006

After doing some research, we decided to add the line set consistent_coloring=2 in the file /etc/system. Basically, this parameter tunes the use different page placement policies on the UltraSPARC. For more details see Solaris Kernel Tunable Parameters.

April 10, 2006

We discovered a group in that was doing benchmarking on T2000. The are in RWTH Aachen University, Germany.

April 9, 2006

We tried to install the packages of LAM/MPI but we were not successful.

April 6, 2006

We obtained the first results from our shared memory application. The code used was Tpetra from Sandia National Laboratories. We were surprised by the fact that the performance scaled up to 16 threads and then went down. For more details see the report of the experiment or go to results section.

April 4, 2006

We run the first program using shared memory on T2000. The standard used was OpenMP. The code was the following:

 
#include <stdio.h>
#include <omp.h>
int main(void){
	int tid; omp_set_dynamic(0);
	omp_set_num_threads(10);
	#pragma omp parallel private(tid){
		/* Obtain thread ID. */
		tid = omp_get_thread_num();
		/* Print thread ID. */
		printf ("Hello World from thread = %d\n", tid);
	}
}

We found this code in the site called "OpenMP Support in Sun Studio Compilers and Tools" (see Appendix A.1).

March 24, 2006

We downloaded and installed Sun Studio 11. The installations was direct and we did not have any problem.

March 22, 2006

We discovered the tools offered by Sun to develop parallel applications. You can find them here.

March 20, 2006

We created users accounts and set the IP addresses. We started to think about the benchmarking code.

March 16, 2006

We get Solaris 10 up and running. The startup was easy and without problems. The documentation was clear. Cool!

March 14, 2006

We received the machine! It was an exiting moment and we started to work on it immediately. We plug a serial terminal and try to browse the starting system.