Enabling code for HPC through Advanced User Support
Written by Sabry Razick
The Optimization of PMetrics is an Advanced User Support (AUS) project which successfully achieved the stated objectives in the project plan, allowing for larger and more complex pharmacokinetic models to be run on the HPC infrastructure.
Five stages towards success
- During the first stage, wrappers were developed for the software to be executed in headless-batch mode.
- In the second stage, the pipeline was modified to accept larger Input data, which was the main limitation faced by the requesters
- In the third stage the processing time was reduced by using Intel® Math Kernel Library and making parallel execution possible using OpenMP technology
- In the fourth stage, scalability was evaluated and optimal resource allocation was decided for the Saga computer cluster.
- An overall 15 times speed improvement over the initial previous version was recorded in the final version.
The AUS has provided the end users with improved infrastructure, shifting the workload from personal workstations and laptops to an HPC cluster, which the following user experiences confirm.
"The AUS has been of great value to our research group, and has allowed us to overcome obstacles that we previously couldn’t. We are now able to run multiple, more complex models at a HPC, which is a drastic upgrade to our infrastructure".
"Overall we are very grateful for the results from this AUS, which as previously mentioned has drastically increased our research capabilities due to improved infrastructure."
- Markus Herberg Hovd - M.Sc. Pharm, Department of Pharmacy, University of Oslo
Interested in AUS and the final report?
For detailed information about this advanced user support, please download the AUS Final Report Optimization of PMetrics.
This work was carried out over a little more than 4 months and with 1 PM's effort. If you are interested in advanced support from Sigma2, you can find more information on the Sigma2 advanced user support service page.
The original PMetrics project page (external)