How Community Software Ecosystems Can Unlock the Potential of Exascale Computing, published in Nature Computational Science in Feb 2021, talks about the need to embrace software ecosystems as first-class citizens.
|Paper title||How community software ecosystems can unlock the potential of exascale computing|
|Authors||Lois Curfman McInnes, Michael Heroux, Erik Draeger, Andrew Siegel, Susan Coghlan and Katie Antypas|
|Publication||Year 2021, Nature Computational Science Journal, DOI: 10.1038/s43588-021-00033-y|
How Community Software Ecosystems Can Unlock the Potential of Exascale Computing is an article published in the journal Nature Computational Science in Feb 2021. This article re-affirms the need to acknowledge the important role of software as the foundation of sustained collaboration and scientific progress, in order to exploit the capabilities of emerging exascale computing systems. Developing such software would require a wide range of expertise emanating from diverse fields of applied mathematics, computer science, and specific domain sciences. Keeping this in mind, the article states that one needs to look beyond casual alliances of applications and software teams for such cutting-edge systems and focus on encouraging exploration of collections of interdependent products whose development teams have incentives to collaborate to provide aggregate value, where the whole is greater than the sum of its parts.
Community software ecosystem perspectives are essential in the exascale era, chiefly due to disruptive changes in computing architectures, increasing system heterogeneity, increasing number of diverse users, increased software-centric collaborations across disciplines, having new science drivers in HPC such as Artificial intelligence, and ultimately having to deal with more foundational topics of software dependency management, reusable software products, software quality assurance, etc. Therefore, close involvement of application teams with software community ecosystems - not just as passive consumers but more as active collaborators - is essential to achieve overall science goals in the exascale era.
The paper is an interesting read and emphasizes that we can all play important roles in addressing several of the technical, cognitive, and social challenges in scientific software by catalyzing change in our own projects, institutions, and communities.