EL3

Table Of Contents

  1. 1. What is EL3
  2. 2. Uploading data

1.  What is EL3


EL3 (Elab's Elbas Elaborator) is a tool designed to analyze HPC and GRID benchmarks. It can perform a variety of diffrent analysis generating charts and tables, using menus and checkbox to refine the data you want.

EL3 is in beta stage and can be found here: http://el3.escience-lab.org/.

2.  Uploading data


Step 1: create and upload a file for the platform you will run elbas on. This file has to follw exactly the same syntax (inclusing spaces) of the following example:

 <sysinfo>
  <sys_id>cl-dell</sys_id>
  <cpu_type>Intel Xeon X7460</cpu_type>
  <cpu_mhz>2660</cpu_mhz>
  <cpu_cache>9216 KB</cpu_cache>
  <cpu_cache_l3>16384 KB</cpu_cache_l3>
  <fsb>1066</fsb>
  <cpus>24</cpus>
  <ram_total>50331648k</ram_total>
  <ram_used>-1k</ram_used>
  <kernel>Linux 2.6.27.4-server-2mnb #1 SMP Thu Nov 6 13:49:28 EST 2008 x86_64</kernel>
  <distribution>Mandriva Linux 2009.0</distribution>
  <network>n.a.</network>
  <type>HPC</type>
  <is_cluster>0</is_cluster>
  <cpus_per_node>-1</cpus_per_node>
  <cores_per_chip>6</cores_per_chip>
 </sysinfo>

Step 2: create and upload a file for the benchmark you will run. Some parameters are standard (i.e. Code name, dataset..) and some depends on your softwar stack (libraries, compiler...)

<benchmark>
 <benchmark_id>gromacs-hpc-openmpi-1.3</benchmark_id>
 <code>Gromacs</code>
 <bench_and_dataset>grobench-1</bench_and_dataset>
 <time_internal_description>n.a.</time_internal_description>
 <BLAS_libraries>MKL 10.0.5.025</BLAS_libraries>
 <LAPACK_libraries>MKL 10.0.5.025</LAPACK_libraries>
 <FFT_libraries>FFTW3 3.2</FFT_libraries>
 <MPI_libraries>Openmpi 1.2.3</MPI_libraries>
 <OTHER_libraries></OTHER_libraries>
 <compiler>Intel Fortran Compiler 10.1.018</compiler>
 <architecture>64</architecture>
 <code_version>4.0.2</code_version>
 <affinity>0</affinity>
 <notes></notes>
</benchmark>

Step 3: upload xml files generated by elbas. You don't need to know anything about these files but to explain how the system work here is an example:

<results>
 <run>
  <benchmark_id>pwscf1-grid-0.1-32</benchmark_id>
  <n_cpus>1</n_cpus>
  <time_real>458.26</time_real>
  <time_user>436.78</time_user>
  <time_sys>21.25</time_sys>
  <time_internal>381.75</time_internal>
  <output> .... plain text output .... </output>
 </run>

 <run>
  <benchmark_id>pwscf1-grid-0.1-32</benchmark_id>
  <n_cpus>1</n_cpus>
  <time_real>458.26</time_real>
  <time_user>436.78</time_user>
  <time_sys>21.25</time_sys>
  <time_internal>381.75</time_internal>
  <output> .... plain text output .... </output>
 </run>

</results>