Willing to help and eager to learn. With this state of mind, John Cazes deputy, director of the HPC group at the Texas Advanced Computing Center (TACC) and Nick Thorne (Senior Systems Administrator) at the University of Texas at Austin, arrived in Portugal. As the clock ticks to have Deucalion, the new Portuguese supercomputer, fully operational, the two TACC experts with extensive experience in supercomputing praised Deucalion’s architecture, emphasizing the ARM processor — suitable, for example, for machine learning techniques and protocols.
This is not the first time that John Cazes has collaborated directly with Portuguese teams. His most recent collaboration happened under the BigHPC project, one of the strategic projects sponsored by the UT Austin Portugal Program, which has rolled out 15 papers so far. “The focus of the work was to create an environment to support users through containers, which will be a big part of the future – since much of the software used in machine learning is made available through containers”, Cazes recalled.
In the past, cooperation between the UT Austin Portugal Program and TACC has proved essential to Portugal’s unprecedented advances in HPC research and innovation. One of the most concrete examples concerns the creation of the Minho Advanced Computing Centre (MACC) and the installation of the BOB supercomputer – the first Supercomputer installed in Portugal. This machine, donated by TACC to Portugal through the Program, paved the way for Portugal’s successful application to host the EuroHCP Deucalion, a green supercomputer capable of peak performance of 10 Petaflops or 10 million billion calculations per second.
”There are no right answers. Regarding system administration, we’re facing a time of change, with much of the systems’ performance coming from GPUs. This substantially benefits Artificial Intelligence (AI) workloads, for example. We’re also at a transition point from direct liquid cooling to CPU sockets and GPUs, which will also bring impactful changes when it comes to systems”, said the researcher.
Despite continuous evolution and the lack of a magic formula, the two TACC experts have no doubts about the unavoidable role that supercomputing will play in the future, especially in the light of machine learning – which requires the same hardware. ”Supercomputing has traditionally been used for simulations in physics and chemistry, experiments that can’t be done in a laboratory. Only on supercomputers”, points out John Cazes.
Advanced Computing is one of the core scientific areas of the UT Austin Portugal Program in Phase 3. It promotes and supports joint projects and training actions on high-performance and high-throughput computing systems, quantum computing, data management and visualization, aiming at better exploiting the use of advanced computing facilities, namely those offered by the Program through the Texas Advanced Computing Center and the Minho Advanced Computing Centre, in Portugal.
