IT-4906/IT/LT CPU benchmark¶
Requirements¶
The CPU performance tests shall be conducted on the proposed CPU server that will be given to CERN as a sample. HEPScore23 and all of its configuration and result files shall remain on the sample server(s) if and when handed over to CERN.
All tests shall be performed with the required RHEL 9.3 installation. The results of the HS23 measurements shall be reported to CERN in the technical questionnaire (Annex A). In addition, an archive of all the produced files of the benchmark shall be created and provided to CERN (see point 9. below).
Running the HS23 benchmark¶
- Make sure all the required packages described in the RHEL 9.3 installation page are available.
- The benchmark has an additional set of software requirements:
- Container engine Apptainer version 1.1.6 or higher;
- Python version 3.9 or higher;
- python3-pip;
- Create the directory
HS23
under/pool/
. - Download the script
run_HS23.sh
available at this link. This script contains all the configurations which are required to run the benchmark. - Edit the script and in particular the
SITE
andPURPOSE
fields. It is important to specify the name of the Bidder (more info can be found insiderun_HS23.sh
). - Make sure that the script is executable by doing
chmod +x run_HS23.sh
.
Info
Please notice that the benchmark will take several hours to complete, it is suggested to launch it from a tmux
or screen
session
-
Run the benchmark like this:
./run_HS23.sh
. -
Once it is finished, the results will be found in the
/pool/HS23/bmk_run/run*
folders. - The result of the run(s) can be quickly filtered out for example running
grep -r "Final result"
from/pool/HS23/bmk_run
- IMPORTANT: it is mandatory to create an archive of the
/pool/HS23/bmk_run/run_*
folders and provide its full content to CERN, with the exception of theunpack
subfolder which is not necessary to include. As an example, the following commands generate an archive:
cd /pool/HS23
tar -czf hs23.tar.gz --exclude=./bmk_run/run*/HEPSCORE/unpack ./bmk_run
It is important to make sure that enough space is available to host the benchmark image, its produced files and the final results. By default the script assumes that the benchmark image is in /pool/HS23/
and the rest of the data in the same location from where the script is launched. It is permitted to adapt accordingly the script (and the instructions above) to point to a different mount with enough space.