US NSF - Dear Colleague Letter: Cyberinfrastructure Framework for the 21st Century - A Vision and Strategy for Software for Science, Engineering, and Education (nsf12113)

NSF 12-113

Dear Colleague Letter: Cyberinfrastructure Framework for the 21st Century - A Vision and Strategy for Software for Science, Engineering, and Education

NSF will take a leadership role in providing software as enabling infrastructure for science and engineering research and education, and in promoting software as a principal component of its comprehensive CIF21 vision. This includes ensuring comprehensive, usable, and secure software and services to further new scientific discovery and innovative education approaches by its researchers working in a globally connected and data-enabled world; fostering sustainable communities of software users, researchers, developers, industrial scientists and engineers, educators, and students that span disciplines, professions, and regions/countries; and promoting new approaches to learning and workforce development in software, and supporting investigations in the use of software for novel learning mechanisms. Reducing the complexity of software will be a unifying theme across the CIF21 vision, advancing both the use and development of new software and promoting the ubiquitous integration of scientific software across all disciplines, in education, and in industry.

Introduction and Background

Software is a critical and pervasive component of cyberinfrastructure for science, engineering, and education. Software is essential at all levels, providing the low-level drivers and transport protocols for networks; operating and file systems for scientific computing; runtime or development environments for high performance and/or distributed computing; new collaborative environments for virtual organizations; improved and coupled modeling, simulation, design, and visualization capabilities; as well as expressing the complex algorithms used to model and then analyze and understand data and processes in science and engineering.

Software is fundamentally computer code. It can be delivered to end users in multiple formats, ranging from an archive that a user downloads and builds to an executable or a service running on a remote system to which a user connects. Especially at large scale, software is generally difficult to design, implement and then maintain, and the software needed by the science, engineering, and education communities is particularly complex. Software must be reliable, robust, and secure; able to produce trustable and reproducible scientific results; yet its architecture must be flexible enough to easily incorporate new scientific algorithms, new capabilities, and new opportunities provided by emerging technologies. Software also must be supported, maintained, developed and eventually replaced in part or in entirety, over its lifecycle.

As computation and data analysis become increasingly important in scientific research, education, and industrial innovation, "enabling software and systems are needed to create an environment in which the barrier to access is low for innovation and new discovery," as stated in the PCAST report Designing a Digital Future. Software needs to transition from a set of individual research projects to a production infrastructure.

The recent task force report of the Advisory Committee on Cyberinfrastructure [1] identifies several key challenges in software covering compute-intensive science, data, software evolution, and institutional barriers to software. Key recommendations from the task force include providing long term, comprehensive support for software at different levels; addressing verification, validation, uncertainty quantification, sustainability and reproducibility; providing policies for open source software; coordinating software activities across NSF, with other federal agencies, and with industry; and mechanisms for community input on software priorities. Improving education and training across disciplines in software use and development is also a recognized challenge.

This Strategic Vision identifies priority areas for NSF investment that will facilitate important and tangible progress in moving 21st Century science, engineering, and education toward more effective use of software and services, treating software as a principal component of cyberinfrastructure, and recognizing that the complexity of software has historically been underappreciated. The NSF vision is to facilitate software infrastructure that works easily at scale, encourages reuse, and efficiently promotes innovation while retaining reliability. This vision promotes greater balance in priorities, coordination, and leveraging, and encourages new strategies for fulfilling the maximal potential of prior cyberinfrastructure investments and new NSF investments.

Goals for Software for Science, Engineering, and Education

To meet the challenges before it, NSF will adopt, as part of its larger CIF21 mission and program, five strategic goals for delivering and sustaining software to advance science and engineering research and education:

Capabilities: Support the creation and maintenance of an innovative, integrated, reliable, sustainable and accessible software ecosystem providing new capabilities that advance and accelerate scientific inquiry and application at unprecedented complexity and scale.
Research: Support the foundational research necessary to continue to efficiently advance scientific software, responding to new technological, algorithmic, and scientific advances.
Science: Enable transformative, interdisciplinary, collaborative, science and engineering research and education through the use of advanced software and services.
Education: Develop a next generation diverse workforce of scientists and engineers equipped with essential skills to use and develop software. Further, ensure that the software and services are effectively used in both the research and education process realizing new opportunities for teaching and outreach.
Policy: Transform practice through new policies for software addressing challenges of academic culture, open dissemination and use, reproducibility and trust of data/models/simulation, curation and sustainability, and that address issues of governance, citation, stewardship, and attribution of software authorship.

Figure 1: Software is an integrated component of the overall coordinated CIF21 framework.

Strategies for Software for Science, Engineering, and Education

The NSF Software for Science, Engineering, and Education strategy plan is part of the overall, coordinated CIF21 framework (see Figure 1) in which software fulfills the potential of new advanced computing infrastructure; provides new data infrastructure; bridges education and cyberinfrastructure between campuses, cities, towns, and countries; advances and expresses computational and data-enabled science and engineering; and enables a new generation of scientific and engineering communities.

To meet each of the software goals, NSF has developed a set of strategies.

Support the creation and maintenance of an innovative, integrated, reliable, sustainable and accessible ecosystem of software and services that advances scientific inquiry and application at unprecedented complexity and scale.

The software ecosystem must catalyze and support emerging, new thinking, paradigms, and practices for science that are fundamentally interdisciplinary and data-driven, and that integrate as an enabling layer with all other CIF21 activities. The software cyberinfrastructure must help researchers and application scientists and engineers address problems of unprecedented complexity, scale, resolution, and accuracy by integrating computation, data, networking, observations, experiments, and disciplines in novel ways. Previous and ongoing investments in software across the Foundation need to be leveraged and connected to a well-coordinated software plan that addresses sustainability. Evaluation of software efforts will be an important activity, tracking and monitoring both the quality of software and the scientific impact.

NSF will address software though a tiered approach that is embedded throughout the Foundation's research programs, building capabilities through different activities for

Software Elements: targeting small groups that will create and deploy robust software elements for which there is a demonstrated need that will advance one or more significant areas of science and engineering.

Software Frameworks: targeting larger, interdisciplinary teams organized around the development and application of common software infrastructure aimed at solving common research and industrial problems, resulting in sustainable community software frameworks serving a diverse communities.

Software Institutes: establishing long-term hubs of excellence in software infrastructure and technologies, research and application communities of substantial size and disciplinary breadth.

Reuse mechanisms: Incentivizing individuals and communities to use and build on existing infrastructure to advance science and engineering.

NSF will provide long-term investment in activities that:

Sustain and advance software infrastructure and software development at different tiers, from individual disciplinary research groups innovating new software elements to community-scale tools and frameworks to large-scale institutes coordinating broad activities across multiple disciplinary areas.

Address all parts of the software lifecycle, from transitioning new research into practice, to operating and supporting well-used software, to ceasing support for superseded or lesser-used software.

Recognize that software strategies must include the secure and reliable deployment and operation of services, for example by campuses or national facilities or industry, where identity, authentication, authorization and assurance are crucial operational capabilities.

Provide mechanisms to incorporate innovations in software, and for software to rapidly evolve to leverage advances in technology and new functionality required by scientific disciplines.

Result in high-quality, usable, secure, vulnerability-free, sustainable, robust, well-tested, and maintainable/evolvable software; and which promotes the sustainability of solid and useful on-going investments.

Ensure that software is well documented, disseminated and discoverable, with accessible resources for training and user ratings/reviews.

Promote the use of elements of the software infrastructure in learning and workforce development activities.

Create generic models of both the cost of making software reusable and the benefit of software reuse, and use specific instances of these models for software decisions.

Support the foundational research necessary to continue to efficiently advance scientific software, responding to new technological, algorithmic and scientific advances.

Core programs at NSF already support areas of foundational research to advance software development. Clear pathways must be in place for such foundational research to impact operational software and services, so key gaps in research can be identified and resolved.

NSF will invest in:

Activities that determine gaps between current software and current and future science, engineering, and education needs.

Research activities that advance software development, use, and accessibility, including:

Programming paradigms that address the use of massively parallel computers, highly distributed computers systems (including private, public, and hybrid clouds), complex file systems (both parallel and distributed), new accelerator architectures and the potentially hybrid systems that will be built from them;

High-level abstractions and frameworks that promote code reuse and sharing, model extensibility and interoperability, and simplify domain specific programming;

Middleware for dynamic data-driven workflows;

Software for collaborative science and engineering (technologies for teams, data, and computing);

Paradigms for verification, validation, uncertainty quantification, assurance, provenance that will ensure trustworthy and reproducible scientific and engineering findings;

Resilience/fault-tolerance: algorithms and techniques for avoiding, discovering, and recovering from anomalous conditions;

Tools and services for gateways/portals/hubs;

Associated software productivity tools, addressing for example testing, debugging, profiling and visualization;

Interfaces and integration between heterogeneous entities: imaging, devices and systems, software, visualization and data mining, and control and decision making;

Goal-based (autonomous and adaptive) software and services.

Understanding issues that prevent reuse; and incentivizing reuse in the development process.

Developing pathways to practice that move the outputs of research activities into the infrastructure, advertise them, document them, and train users in academia and industry.

Encourage the transition of mature software into industrial and medical practice to lead innovation and product development.

Developing principles and practices for sunsetting existing software elements, and possibly replacing them with new ones, with the participation and awareness of the user and developer communities.

Enable transformative, interdisciplinary, collaborative, science and engineering research and education through the use of advanced software and services.

Broad investments need to be made to support the use, and associated advancement, of production software for all scales of science problems where software integrates the use of new algorithms and scientific capabilities with cyberinfrastructure resources. This software must be driven by the requirements of scientists and engineers, considering both immediate needs and longer-term goals.

NSF will invest in activities that:

Develop an NSF-wide program in computational and data-enabled science and engineering (CDS&E) to provide the research and innovation in new methods, algorithms, and approaches that will advance software for science and engineering.

Address a range of research questions, from grand-challenges to long-tail research, by engaging a comprehensive and integrated approach to science and engineering, utilizing software, high-end computing, data, networking, facilities, and multidisciplinary expertise.

Engage early-stage researchers in the use of complex and visionary end-to-end scientific use cases in different disciplines that drive innovation in cyberinfrastructure development and use.

Support community-building activities to promote collaboration that crosses disciplinary, institutional, and geographic boundaries through shared software concepts and standards.

Develop a next generation diverse workforce of scientists and engineers equipped with essential skills to use and develop software. Further, ensure that the software and services are effectively used in both the research and education process realizing new opportunities for teaching and outreach.

NSF must integrate software use with educational activities at all levels to prepare students for 21st Century careers and to maintain competitiveness in the international marketplace and create and support career paths for software developers.

NSF will invest in activities that:

Provide students with the expertise to use and extend the most up-to-date tools and techniques and contribute to the scientific software infrastructure,

Provide new curricula and methods to teach students at all levels and across all disciplines best practices for software engineering,

Provide general curricula that enable software appreciation and literacy among the public and policy making communities,

Facilitate and encourage professional career tracks in computational science and software engineering.

Encourage proactively broadening participation of women and underrepresented minorities.

Transform practice through new policies for software addressing challenges of academic culture, open dissemination and use, reproducibility and trust of data/models/simulation, curation and sustainability, and that address issues of governance, citation, stewardship, and attribution of software authorship.

Sustainable models for software infrastructure development will require the evolution of existing policies as well as social change in the research and academic communities.

NSF will lead initiatives targeted at evolution and change in the following areas:

Widespread adoption of open source models for software development and dissemination that include high quality documentation and accepted engineering practices, leading to software that is accessible, understandable and reusable;

Mechanisms for citation of software as distinct products of scholarship, promoting standards of academic credit and rigor for software;

Cost-effective strategies for broad, interdisciplinary collaboration in software development, including with international and industrial entities;

Governance and sustainability models, including those currently in use by open source communities and by industry;

Development and use of metrics that measure software usage and impact on science, engineering and education;

Determining existing barriers to software reuse, from both the developer and user points-of-view, and promoting mechanisms to overcome them.