# Lev Lafayette, 2020. Contributing To the International HPC Certification Forum. Presentation to eResearchAustralasia, October 23 https://conference.eresearch.edu.au/2020/09/contributing-to-the-international-hpc-certification-forum/ There are two major issues confronting research computing today. The first is the fact that data size and complexity is growing much faster [1][2] than the capacity of personal computational devices to process that data, primarily to the heat issues associated with increased clock-speed (Dennard scaling) [3]. There have been numerous hardware and software attempts to reduce this problem; multi-processor systems, multi-core processors [4], shared and distributed memory parallel programming [5], general purpose graphics processing units [6], non-volatile memory (the PDP-11 gets the last laugh), RoCE/RDMA and network topologies, and so forth - all of which are providing improvements, but not solutions to this constant chase and irresolvable conflict between data acquisition and data processing. However this presentation not about these raw hardware capabilities or software optimisations, but rather the very real wetware issue - human ability to use such tools. Ignoring this is a critical issue in the high performance computing (HPC) space [7], as HPC systems are the best tool for processing large and complex datasets and as a result are increasingly in demand. Over the decades however, personal computational devices have concentrated on developing high-level Graphic User Interfaces (GUIs), in direct contradiction to the needs of computation [8]. Indeed the very desire to "dumb down the interface" rather than "skill up the user" in HPC [9] is fraught with three major intractable problems: Firstly, such "user-friendly" interfaces are often subject to significant disruptive changes that require re-skilling, whereas core command-line tools have remained largely constant for thirty years and more. Secondly, GUI systems by necessity must involve a loss of computational power and fine-tuning, as behind every GUI executable icon there is a lower-level shell and kernel commands. Thirdly, the GUI reduces understanding of how the system operates, which is necessary to maximise computational performance. Unfortunately, the formal education system has not yet responded sufficiently to this pressure with some notable exceptions such as the University of Waikato in New Zealand [10], leaving HPC centres to provide basic training. Of course, HPC system engineers are not advanced adult educators (usually), and they do not necessarily have a grasp of the pedagogical-andragogical continuum, proximal development, connectivism, structured and integrated teaching techniques and so forth. Nevertheless, they do understand learning objectives, and it is from this approach the international HPC Certification Forum was established in 2018 [11], developing from the Performance Conscious HPC (PeCoH) project in 2017 from the Hamburg HPC Competence Center (HHCC), which had the explicit goal of creating the broad standards for an "HPC driving license" [12]. Since its establishment, the Forum has developed a detailed skill-tree across multiple branches (e.g., HPC Knowledge, HPC Use, Performance Engineering etc) and levels of competencies (basic, intermediate, expert) where very specific skills have particular competencies. In addition, the Forum has developed a summative examination system and a PGP-signed certificate. "The goal of the HPC Certification Forum is to clearly categorize, define, and examine competencies expected from proficient HPC practitioners" [13]. This is a stated separation between the examination and certification, which the Forum does govern, from curriculum development and content delivery, which is left to HPC education providers. But apart from a necessary organisational independence, this also provides a requisite feedback mechanism for HPC education providers. The HPC Certification Forum is relatively small, with approximately a score of participants; there are thousands of HPC sites across the globe, and perhaps millions of HPC users. It is not possible in these circumstances for the Forum to guarantee providing the relevant and appropriate examinable content despite heterogeneity. The Forum requires an eco-system for development, and to keep up-to-date with new technologies, meaning that the education providers will actually drive the content through formal and informal feedback. For its part, the Forum provides endorsements, so that education and training providers as HPC CF Endorsed Training. The HPC CF exams are expressed as a tree diagram with branches for broad areas as structured content (e.g., software development, performance engineering, HPC knowledge, HPC use) with skills in these areas defined as a set of learning outcomes, with multiple levels (basic, intermediate, and advanced, a form of "scaffolding" in education). This does mean that content and delivery providers can treat the examinable content as a curriculum with scaffolding. Likewise content-providers can propose multiple-choice questions within the Forum's examinable learning objectives, along with allowing for any errors or emissions to be identified and fixed. So the eco-system provides an approach where the Forum's material that those who provide curriculum and content can draw upon and seek endorsement, but at the same time have the opportunity to contribute to the learning objectives and even the questions that make up the requirements for the HPC Certification. These provide real avenues, in addition to more typical actions such as joining the HPC Certification Forum mailing list or even, for those with time and energy, joining the governance committee, that those involved in HPC education and training can contribute in a very real sense to the Certification programme and help ensure its success. References [1] Hilbert M, López P. "The world's technological capacity to store, communicate, and compute information". Science. 332 (6025): 60–5, 2011 [2] Guo, Huadong, et al. "Scientific big data and digital earth." Chinese Science Bulletin 59.35 (2014): 5066-5073. [3] Markov, Igor L. "Limits on fundamental limits to computation." Nature 512.7513 (2014): 147 [4] Hruska, J., The death of CPU scaling: From one core to many — and why we’re still stuck, extremetech.com, February 1, 2012 http://www.extremetech.com/computing/116561-the-death-of-cpu-scaling-from-one-core-to-many-and-why-were-still-stuck [5] Sutter, Herb. The Free Lunch Is Over: A Fundamental Turn Toward Concurrency in Software Dr. Dobb's Journal, 30(3), March 2005 [6] Tsutsui, Shigeyoshi., Collet, Pierre (eds), Massively Parallel Evolutionary Computation on GPGPUs, Springer-Verlag, 2013 [7] Wilson, Greg. High Performance Computing Considered Harmful, 22nd International Symposium on High Performance Computing Systems and Applications, 2008 [8] Reiser, Martin. The Oberon System User Guide and Programmer's Manual. ACM Press, 1991 [9] Lafayette, Lev. Software Tools Compared To User Education in High Performance Computing. Proceedings of The Higher Education Agenda conference, Gold Coast, May, 2015 [10] Lane, Jo. Scientific supercomputing: Teaching practical skills for credit, Presentation to eResearchNZ, Dunedin, Feb 14, 2020 [11] Himstedt, Kai., et al., An HPC Certification Program Proposal Meeting HPC Users' Varied Backgrounds, 2018 [12] Kunkel, Julian., et al,, Towards an HPC Certification Program, Journal of Computer Science Education, Volume 10, Issue 1, 2019, p88- 89 [13] Kunkel, Julian et al., "The HPC Certification Forum: Toward a Globally Acknowledged HPC Certification" in Computing in Science & Engineering, vol. 22, no. 04, pp. 110-114, 2020. doi: 10.1109/MCSE.2020.2996073