About the project

Somebody Has to Go First

Stakeholders agreed that introductory computing needed modernizing — but nobody had a modernized course to copy. This project builds one, in eight different institutional contexts, and measures whether it works.

Goals

The Problem, Stated Plainly

The CDER Center ran a series of stakeholder workshops from 2020 to 2023, supported by NSF CyberTraining Institute Conceptualization Award #2002649, examining the systemic barriers to broader PDC adoption — among them the continued dominance of a sequential model of computation in early courses. Employer stakeholders — including national labs and agencies — reported that computer science and engineering students are not graduating with a sufficient grasp of the software development process for the modern computing ecosystem, which relies on parallelism, distribution, asynchrony, scaling, integration across disparate libraries and data sources, test-driven design and pervasive concerns for security.

The lack of preparation is rooted in an obsolete algorithmic model of computing — sequential, synchronous, with text-based I/O — that instruction relies on from introductory programming through data structures, algorithms and software engineering. Once that model is entrenched, students may only encounter PDC ad hoc via electives.

Education stakeholders said changing the model is nearly impossible because of a lack of exemplars, few teaching materials, little instructor training in modern computing, and little evidence that changing the model benefits students. Perhaps the greatest impediment is that academic stakeholders find it difficult to visualize what a different approach would look like — and most feel such a change is simply not possible. What the workshops heard repeatedly was: “someone else needs to go first.”

Read this in full — Chapter 0 § 0.1, p. 9 →

1Build a modern first-year sequence around a computing model that includes parallelism and distribution as found in current systems.
2Generalize it across two colleges and two languages, to show the model is not language- or context-specific.
3Evaluate with pre/post instruments grounded in education science, measuring unmodified courses first as a control.
4Transfer to six further institutions chosen for diversity of size, schedule and student population, and find what makes it widely usable.
5Disseminate everything — course sequences, materials, chapters and results — through active outreach.

Scope

Three PDC Topics, Chosen Deliberately

Not everything in the curriculum guideline belongs in a first-year course. The project narrowed to three ideas that attach naturally to existing CS1 and CS2 content.

Data parallelism

Attaches to loops, arrays, searching and sorting. A range-based for loop already models dependence-free processing of a collection — a candidate for parallelism sitting in plain sight.

Distributed data access

Modern file and user I/O APIs read and write serialized objects, even from remote repositories. Students pull live data from a real remote service instead of parsing console text.

Event-driven processing

Graphical interfaces exemplify event-driven processing, which illustrates many of the basic ideas behind control parallelism — and is the paradigm internet-native students already inhabit.

How it was run

Three Sets of Teams, Working Closely Together

Development

Two institutions, two languages

Tennessee Tech (C++, semester system, large lecture and lab CS1) and Knox College (Java, quarter system, classes of about 24) each created exemplars for their first-year sequence, coordinating so the same PDC concepts were covered in context-specific ways.

Testing

Six adopting institutions

Recruited through the CDER community and the Edu* and SIGCSE networks, then selected to balance language and institution type. Each transferred the exemplars into their own courses, identified opportunities and challenges, and fed the experience back to improve both sets of teams.

Backbone

Coordination & expertise

Prasad, Sussman, Thota, Vaidyanathan and Weems ran the training workshops and biweekly meetings, facilitated discussion, and provided PDC domain expertise.

Project timeline

Four years, from recruitment to publication

2023
Recruitment and kickoff

Six testing teams recruited and selected on course fit, institutional willingness and language, then brought together at an in-person summer workshop.
2024
CS1 transfer and baselines

CS1 exemplars transferred to the testing teams. IRB approval obtained, research questions agreed, and baseline offerings taught and measured with no intervention.
2025
CS2 work and intervention runs

Baseline data and the first experimental CS1 runs presented; focus turned to CS2 while the new courses ran and evidence accumulated.
2026–27
Analysis and publication

Cross-institution analysis, this e-book and conference reports — plus a follow-on NSF award extending training to AI and Big Data.

NSF support

Award #2321015

CyberTraining: Implementation: Medium

Modern Course Exemplars infused with Parallel and Distributed Computing for the Introductory Computing Course Sequence (2023–2027). The project adopts a “go first” effort to create exemplars of first-year course sequence changes and accompanying materials, while demonstrating their effectiveness and transferability.

Where the need was identified

NSF CyberTraining Institute Conceptualization Award #2002649 supported the 2020–23 stakeholder workshops that examined the systemic barriers to PDC adoption and produced the “someone needs to go first” conclusion this project answers.

Prior support that made this possible

The earlier CDER center grant (curriculum guidelines, dissemination, early adoption) and an NSF CyberTraining implementation grant for broader instructor training in incorporating PDC into undergraduate CS and CE curricula.

Foundational grants

IIS 1143533 · CCF 1135124 · CCF 1048711 · CNS 0950432 — supporting the NSF/IEEE-TCPP curriculum initiative, with additional support from IEEE TCPP, Intel, NVIDIA and IBM.

What comes next

A new NSF award will organize additional instructor training workshops to further infuse AI and Big Data, together with PDC, into CS1, CS2 and Computer Systems courses through the CDER center.

Prior work

This Did Not Arise in Isolation

The CDER Center has worked for more than a decade to integrate PDC concepts into undergraduate computing curricula. Hover or tab through any row for the detail.

Began with a single question: what should every computing undergraduate know about parallel and distributed computing? The resulting guidelines identified PDC topics across programming, architecture, algorithms and cross-cutting areas, and mapped them to levels of coverage. Version I was released in 2012; Version IIβ in 2020, with expanded attention to Big Data, energy and distributed computing. ACM/IEEE CS2013 linked to it explicitly, and CS2023 includes a dedicated PDC knowledge area citing the Version IIβ effort.
Curriculum guidelines alone do not change courses. The CDER early adopter competitions incentivized instructors to incorporate PDC into existing courses and report their experiences. More than 140 early adopters have been supported, and the broader community of early adopters and trained instructors has grown to roughly 300 participants or institutions nationally and internationally. They demonstrated PDC could be infused widely — and revealed a persistent need for complete, coherent, classroom-tested course exemplars, especially for early courses.
The first two volumes, published by Morgan Kaufmann and Springer, gathered chapters on introductory parallelism, threads, fork-join parallelism, performance, scalability, energy efficiency, MapReduce, GPU computing, mobile concurrency and integrative GUI applications, available as free preprints and collectively reaching tens of thousands of downloads. A third volume, focused on adopter experience and resources, is expected from Springer in late 2026.
EduPar (with IPDPS) began as a venue for early adopters to report experiences and discuss curricular and pedagogical issues. EduHPC (with SC) created a complementary venue connected to the HPC community; EduHiPC (with HiPC, in India) extended the model internationally. Together they transformed isolated adoption efforts into an ongoing professional community.
Distinct from the Edu* research workshops, CDER has organized instructor-facing training: NSF CyberTraining week-long summer workshops, tutorials and special sessions at SIGCSE, training associated with EduHiPC, and hybrid or online events. All of this project’s development and testing team members were past trainees or early adopters — the training pipeline is how the project staffed itself.

Partners

Eight Institutions, Deliberately Unalike

No single model fits all languages, calendars, class sizes, student populations or instructor backgrounds. This project emphasizes multiple pathways for PDC infusion.

Quick comparison of CS1 institutional exemplars for potential adopters (Table 1, Chapter 0).
Chapter	Institutional context	Lang.	Primary PDC infusion style	Good starting point for
2 · TNTECH	Mid-sized public university; large CS1 lecture/lab setting	C++	Unplugged Penny and Flag Maker activities; OpenMP/data-parallel programming; remote-data/API assignment	Large-section CS1 adoption; programming-linked PDC infusion; multi-semester evidence
3 · Knox	Small residential liberal arts college; Java-based introductory sequence	Java	Conceptual and programming-oriented CS1 activities adapted to a small-college context	Liberal-arts setting; smaller classes; Java-based adaptation
4 · USI	Public master’s-level institution; introductory object-oriented CS1 course	Java	Unplugged Flag Maker and Penny activities, plugged-in Flag Maker and Parallel Sort, Greenfoot, and activity-level assessment	Structured activity adoption; evidence-rich unplugged and plugged-in examples
5 · UNL	Large R1 institution; multiple CS1 sections and computing majors	C, Java	Codeless modules, visual and conceptual activities, remote-data ideas, scalable conceptual infusion	Low-code pathway; large-institution adaptation; conceptual modules; online classes
6 · Webster	Private primarily undergraduate institution; small-to-moderate CS1 sections	C++	Animations, visualizations, performance examples, and game/simulation-oriented activities	Visualization-first adoption; low-barrier conceptual entry points
7 · Casper	Two-year public community college; small cohorts and mixed student pathways	C++, Python	Custom digital-textbook material, unplugged activities, OpenMP, remote data, and physical computing	Community-college setting; small cohorts; varied delivery formats
8 · HPU	Private undergraduate-serving institution with small class sizes	Java	Flag Maker, Penny Sorting, lightweight plugged/code-review extension, and low-resource activities	Small-scale first adoption; low-preparation unplugged activities
9 · MSU	Public university; focused CS1 intervention	Java	Flag Maker-centered data-parallelism activity with supporting animations and assessment	Minimal-infusion model; short, focused CS1 intervention

Chapter authors for each institution are listed on the team page.

Search the exemplars Meet the team

Get in touch

Thinking of Adopting?

The material is free and needs no permission to use. Tell us what you are considering and we can point you at the right chapter, help you scope a first adoption, or arrange a short consultation.

Knowing where the exemplars travel also lets the project report to NSF.

Request adoption help Join the Discord

Prefer plain email? Write to contact@nsfexemplar.cdercenter.org. Every route reaches the same people.