A Vision of Computing in 10+ Years Steven R. Brandt Frank Löffler sbrandt@cct.lsu.edu knarf@cct.lsu.edu Center for Computation and Technology & Center for Computation and Technology Department of Computer Science Louisiana State University Louisiana State University Baton Rouge, Louisiana 70803, USA Baton Rouge, Louisiana 70803, USA Once upon a time, single scientists could essentially difficult to find, which underlines the importance of efforts know everything that was known in science, could under- understanding the reasons. We expect that essentially stand the concepts of their time. Their expertise could every scientist in the very near future will need to have stretch from physics, chemistry and biology to medicine, a certain amount of skills also outside of their base art, and even fields of study we would not consider sci- discipline, specifically within domain science, computing, ence today. This picture has drastically changed with the algorithms, and hardware, as well as collaborative software increasing amount of knowledge in each of these fields, development. That burden needs to be minimized, both separating scientists in different fields, often even linguis- by reducing the amount of interdisciplinary knowledge tically. a scientist needs to acquire and in the effort needed to In addition, especially in experimental fields, performing acquire it. science became so complex that single teams or even We do not have an answer, and we believe that no small collaborations of teams are no longer sufficient. general answer exists. However, some approaches have Prime examples of well-known teams are the LHC, or the potential to be helpful: Ligo/Virgo collaboration, with specialization happening 1) We find a way to separate the specification of (a) even within a single, well-defined field of science. This science simulation and computational problems (b) numer- transition will only accelerate during the next 10+ years. ical methods and (c) low-level optimizations Doing this would create a library of physics problems One of the main drivers of this transition is an increase which can be implemented by independent groups of in problem complexity beyond what a single individual or computer scientists and optimized by specialists in hard- team can handle. This is increasingly the case in scientific ware. It would enable better partnerships with industry computing, and here we include both two “extremes”: and academia, and give vendors better targets for new high-performance computing (HPC), and single (or few) hardware and software designs. core applications that are used by a much wider scientific 2) We standardize on the basic infrastructure of com- audience. HPC is leading in the increase in complexity puting, i.e. parameter files, compilation and configuration due to hardware changes, as it always incorporates new tools, performing high performance I/O. hardware developments first, but small-scale computing To the extent that this can be achieved, it will lower the will be eventually be affected also. Moreover, software barrier to sharing and understanding codes and reduction paradigms change. Nobody can answer with certainty that of effort. computers in 10 years will look like, or which programming 3) We find a better way to recognize and reward the languages or concepts will be used. It is not even certain people making the biggest contributions to science. It will what the future of current concepts (OpenCL, CUDA, be essential to find a way to distribute rewards, including OpenACC, ...) will look like in five years; the lifetime of a recognition, within teams that not only span hundreds or cluster. thousands of people, but also science fields, continents and However, hardware and implementation complexity are cultural barriers. Some teams try to do this using author not the only drivers of this transition. The increase in lists that include hundreds of scientists, but this in the end possibilities offered by modern computing infrastructure only shows the gross inability of the current reward system brings with it an increase of model complexity, to the to deal with modern science. Unfortunately, these rewards extent that experts in different science fields work on are tightly coupled to the carriers of the scientists through different parts of a single computational model and its narrow-minded ways of comparing the “scientific value” of, implementation, often without being experts in collabora- for example, faculty candidates or promotion within de- tive software development, or formal training in software partments. The earlier these problems can at least begun development. Despite the obvious shortcomings of such to be attacked, even just by spreading discussions about an approach, successful examples of such teams are not them, the higher will be the pay-off by keeping talented This work is licensed under the CC-BY-4.0 license. individuals within academia that currently just don’t quite fit the reward structure. perlinked to standardized explanatory texts with technol- 4) We need to find a way to make the ever-increasing ogy no more advanced than the hyperlink. Done correctly, complexity of science better understandable. While here all scientific knowledge could be crafted into a single tree we mainly aim to be able to effectively communicate making it possible for one to start with any given scientific between experts of different fields, it should also have a publication and to systematically learn all the concepts beneficial effect on the image of science by the general needed to comprehend the paper. population. To summarize, we expect scientific computing in 10+ Currently, attempting to read a scientific paper that is years to look quite different from today. There will be even slightly outside one’s area of specialization is nearly more and more specialization within teams, with team impossible. This need not be the case, and we hypothesize sizes increasing. In done right, we will also see better, that in ten years we find a better way to integrate our more appropriate scientific reward systems, as well as more corpus of scientific knowledge. emphasis on better “documentation of science”, even for Scientific terms, methods, and notations could all be hy- scientists.