CHPC has made significant progress in ensuring world-class computing services to its user-base, affiliated with international science and technology projects and put human capital development in the fore-front of its planning in the 2013-14 financial year.
CHPC has joined the Square Kilometre Array (SKA) project and forms part of the Science Data Processing (SDP) Consortium of SKA together with Cambridge University, SKA-SA and Daresbury Laboratories.
Organsations wishing to affiliate with the global project sent proposals to the SKA Office for participation. This was followed by SKA presentations of Work Package Consortia that showcased an overview of the composition, people, work concepts and approach to each particular work package. In the SDP Consortium, centre is set to contribute towards three full-time equivalent work packages: exascale prototypes, tiered data delivery and science support. In the financial year ended, CHPC participated in a number these SKA meetings. At the January 2014 meeting, held at the Juelich Supercomputing Center, final prototypes were discussed. The current heterogeneous architecture test-bed provided by CHPC, with a mixture of accelerators, will form part of the prototypes. The drive to enhance HPC astronomy competence in the centre has resulted in significant progress in the building the relationships with the MeerKAT and SKA teams.
Team South Africa scooped overall top honours on 19 June 2013 at the international Student Cluster Challenge in Leipzig, Germany, achieving the highest aggregate points total for all the benchmarks included in the competition (Linpack and the chosen applications) and acing the interview with the judges. South Africa’s resounding success was unsuspected since the country was a first time entrant. The team won over seasoned competitors from USA, China, Germany, UK and Costa Rica.
The Student Cluster Challenge is hosted by the International Supercomputing Conference (ISC) and is an opportunity to showcase student expertise in a friendly yet spirited competition. The Centre for High Performance Computing was the organizer of South African competition. In the challenge, students have a unique opportunity to learn, experience and demonstrate how high-performance computing influences our world and day-to-day learning. Held in collaboration with the High Performance Computing Advisory Council and ISC, the Student Cluster Challenge is designed to introduce the next generation of students to the high performance computing world and community.
One of the students who participated in the national competition, Mathew Cawood, completed his honours degree utilising CHPC resources. He has now been awarded a Master’s studentship and placed in the centre’s the Advance Computer Engineering Lab. The experience in the competition will assist South Africa grow a generation of high performance computing expertise for national economic development and for large projects such as SKA. South Africa will enter as defending champions in June 2014 and hopes to continue flying the flag.
The centre’s computational resources enabled about 60 peer-reviewed publications by users and 55 completions of postgraduate theses in the year. The centre is acknowledged in the published work.
After the successful completion of the pilot Bio-Informatics Service Platform (BSP), an extension of this project was granted by the Department of Science and Technology to increase the level of support for this community.
The centre received 171 new users registration in the year, 1197 helpdesk calls, a system utilization of over 91% was realised, were 22.6 million total CPU hours was provided to 537 active users with 172713 running jobs. The centre achieved a system uptime of approximately 83.8% where 5.2% was scheduled downtime for maintenance.
After the implementation of the new workload manager in the third quarter, there was significant improvement of the resource allocation to the research community. The cluster utilisation now averages about 75% with significant improvement of applications’ efficiency, where all CPU reserved for jobs are utilised.