The Problem
Most Latin American countries have some form of public computing infrastructure for research; however, access to these resources has yet to be democratized. Researchers need to strengthen their data processing capabilities, but they often lack solid training, open learning resources, accessible technical support, and a welcoming peer community to help them transition from using a desktop computer to a cluster. Additionally, national regulations on privacy and data sharing restrict the use of foreign computing infrastructure. Cloud computing is not a solution either, as it typically relies on cash funding, which is always intermittent and rarely lasts beyond one government term.
The Solution
This project seeks to ensure access to high-performance computing (HPC) for national and regional research networks in Latin America. It will create a cloud-based computing platform using JupyterHubs and Notebooks, sustain a new inclusive community of practice with community-led governance, and train researchers in best practices for open data processing and storage, including a Train the Trainer program to multiply impact. This will enable researchers, particularly those in resource-constrained environments, to work remotely within an HPC cluster environment, thereby increasing scalability, reproducibility, and the ability to publish data in accordance with FAIR principles. Interactive computing on powerful machines will serve as an intermediate step to generate demand for traditional HPC, with a high complexity barrier. It will also make it possible to maintain the infrastructure for much longer than could be achieved by renting in the public cloud, providing predictability for long-term tasks such as PhDs or research lines.
Why We Do It
MetaDocencia and UNC Supercómputo are uniquely positioned to bring this project to fruition. They bring together UNC Supercómputo’s 15 years of experience providing public computing services for more than 300 research organizations in Argentina, and MetaDocencia’s track record in community-centered training across Latin America since 2020. The collaboration promises a lasting impact and aims for this project not only to be sustainable but also to become a model for the region. This project is a direct extension of the work supported by UNC Supercómputo and the vision and mission of MetaDocencia, as it addresses the gap in access to HPC infrastructure and enables more Spanish-speaking communities in science and technology to engage in open science and contribute to global knowledge.
The Work Model
This project builds on the successful experience of UNC Supercómputo and MetaDocencia in the Catalyst Project, an initiative that facilitated access to cloud computing infrastructure for researchers in Latin America and Africa. MetaDocencia recruited and trained Latin American research communities, while UNC Supercómputo assisted with infrastructure deployment and initiated its local replication. This project represents a natural evolution for MetaDocencia and UNC Supercómputo, leveraging the lessons learned and deep knowledge of the regional research ecosystem to ensure success and mitigate risks. As part of a detailed plan to implement the Jupyter service on existing infrastructure provided by UNC Supercómputo, alongside community building, training activities, and a research study conducted by MetaDocencia, some key actions include:
- Creation of a JupyterHubs and Notebooks cloud service to lower barriers to HPC for research networks in Latin America.
- Promotion of traditional HPC use where applicable.
- Development of a community of practice around the service, with community participation in its governance, fostering a collaborative and safe environment.
- Comprehensive training in best practices in HPC and open science for researchers in resource-limited environments, including training for trainers to maximize knowledge dissemination.
Expected Impact
UNC Supercómputo has provided computing services for research since 2010. In 2023 alone, at least 273 researchers from more than 40 organizations published over 150 academic papers using its shared resources. MetaDocencia has delivered 94 training editions to over 1,500 researchers from 33 countries, achieving an excellent Net Promoter Score (NPS) of 89%. Additionally, it has developed five successful Train the Trainer programs. These achievements demonstrate MetaDocencia’s capability to recruit and train Latin American research and teaching communities. This project aims to enable more people engaged in science, especially those in organizations with limited access to open data services and cloud computing, to scale their work according to best HPC practices. This will lead to greater reproducibility of their research and the publication of FAIR data. Participants will also be able to adequately train others on these topics and related ones, such as the proper preparation of budgets for central computing services in their grant applications. Along the way, they will have an active and supportive community to accompany them.