Mastering Research Data Storage

Okay, so we’re talking about data, right? In today’s lightning-fast research landscape, data isn’t just an output; it’s the very lifeblood of discovery, innovation, and progress. But here’s the thing: generating data is one challenge, and making sure it stays accessible, secure, and genuinely usable for the long haul? That’s a whole different ballgame. It really requires a thoughtful, strategic approach to storage, a kind of digital infrastructure planning that many might overlook initially. Think of it like building a house; you wouldn’t just throw up walls without a solid foundation, would you? Similarly, our digital data needs that robust framework.

We’re going to dive deep into some genuinely impactful real-world examples today, peeling back the layers to see how various organizations, from academic powerhouses to multinational corporations, have tackled their data storage and management puzzles. You’ll see, it’s not just about buying a bigger hard drive; it’s about architecture, governance, culture, and even a bit of foresight. Let’s get into it.

Cost-efficient, enterprise-level storageTrueNAS is delivered with care by The Esdebe Consultancy.

Monash University’s Large Research Data Store (LaRDS): Building a Digital Fortress for Discovery

Imagine a world-leading university, brimming with brilliant minds conducting groundbreaking research across countless disciplines. From climate modeling to medical breakthroughs, from astrophysics to social sciences, the sheer volume of data being generated is, frankly, mind-boggling. That was precisely the challenge confronting Monash University in Australia way back in 2006. They needed more than just a place to dump files; they required a robust, scalable, and enduring storage solution capable of supporting their ambitious research agenda for years to come. Enter the Large Research Data Store, affectionately known as LaRDS.

Now, LaRDS wasn’t just some off-the-shelf solution. It was a visionary undertaking. This petascale storage system, designed from the ground up, offers thousands upon thousands of terabytes of capacity. Just think about that for a second: we’re talking about petabytes, an astronomical amount of storage that can comfortably house vast datasets, high-resolution images, complex simulations, and all the diverse digital artifacts that modern research produces. It’s essentially a digital fortress, meticulously engineered to provide Monash researchers with a secure, long-term infrastructure for their invaluable data.

One of the most remarkable aspects, and something I genuinely admire, is LaRDS’s commitment to accessibility. It’s not some exclusive club for senior professors. Nope, this phenomenal resource is freely available to all Monash researchers, including postgraduate students. That’s a game-changer, isn’t it? It democratizes access to high-end storage, empowering the next generation of scientists and scholars to manage their data professionally right from the start of their careers. Whether you’re working on a massive genomics project or a longitudinal social study, LaRDS supports various types of research data without discrimination.

But accessibility doesn’t mean a free-for-all, naturally. Security and proper data stewardship are paramount. LaRDS ingeniously incorporates sophisticated access controls, allowing users to precisely manage permissions. This means researchers retain granular control over who can view, edit, or share their data, ensuring its security and facilitating appropriate collaboration. Imagine a scenario where a multidisciplinary team needs to access a shared dataset; LaRDS makes it seamless while still safeguarding individual contributions. It’s like having a highly organized, secure digital library where you dictate who gets which key.

Furthermore, LaRDS doesn’t operate in a vacuum. The team behind it clearly understood the importance of an integrated ecosystem. They didn’t just build a silo; they ensured LaRDS integrates seamlessly with critical university applications. For instance, it works hand-in-glove with tools like the Confluence wiki, allowing researchers to document their work and link directly to stored data. Similarly, its integration with the Sakai virtual research environment really streamlines collaboration, letting teams share data, discussions, and findings within a unified platform. This thoughtful integration minimizes friction, making data management an enabler rather than a cumbersome chore.

What can we really learn from Monash’s LaRDS? Well, for one, it shows us the power of proactive institutional investment in research infrastructure. They didn’t wait for data problems to become crises; they built a solution for the future. It’s also a testament to recognizing the diverse needs of a research community and building a flexible system that serves everyone. From a strategic perspective, offering such a comprehensive, well-integrated, and freely accessible service significantly enhances a university’s research capabilities and reputation. It says, ‘We’re serious about your science, and we’ll give you the tools you need to excel.’ And honestly, that kind of commitment is infectious.

Procter & Gamble’s Data Quality Enhancement: Taming the Global Data Beast

Now, let’s shift gears from academia to the corporate behemoth that is Procter & Gamble (P&G). When you’re a global consumer goods giant, operating across countless brands and markets, your data infrastructure can quickly become a sprawling, complex beast. P&G faced a particularly thorny issue: managing critical master data across a whopping 48 disparate SAP instances. Just picture that – 48 separate systems, each potentially holding slightly different versions of the ‘truth’ for product codes, supplier details, or customer information. It’s a recipe for confusion, inefficiency, and ultimately, significant operational headaches.

This kind of fragmentation isn’t just an IT nuisance; it has real business consequences. Inconsistent data means inaccurate reports, flawed inventory management, supply chain disruptions, and slowed decision-making. You’re trying to make strategic choices, but you can’t trust the data points you’re given. It’s like trying to navigate a ship with a compass that gives a different reading every five minutes.

To combat this, P&G didn’t just patch things up. They implemented a comprehensive data quality software solution. This wasn’t just about cleaning up existing errors, though that was certainly a part of it. This initiative focused on streamlining data management processes from the ground up, establishing greater control and governance over their master data assets. By doing so, they aimed to create a single, reliable source of truth across their vast global operations.

What were the tangible wins from this strategic move? Quite a few, actually:

• Improved Productivity through Automation: Previously, many data-related tasks were manual, labor-intensive, and prone to human error. Automating these processes not only freed up valuable personnel to focus on more strategic work but also drastically reduced the time it took to get accurate data into the hands of decision-makers. Think about how much faster a new product launch can move when all the underlying data is clean and ready.
• Reduced Operational Risks: Fragmented data often leads to data leakage – where sensitive information might exist in uncontrolled environments – and rampant duplication, creating confusion and making compliance a nightmare. By centralizing and standardizing their master data, P&G significantly minimized these risks, strengthening their overall data security posture and making auditing much simpler. This is crucial for a company of their size, especially with ever-tightening regulatory requirements.
• Timely Insights for Management: Before, getting a holistic view of operations was like pulling teeth. Now, with cleaner, more reliable data, management gained timely access to critical health reports and performance metrics. Imagine being able to see a consistent, accurate picture of your supply chain performance or market trends across all regions, enabling faster, more informed strategic decisions. This shift from reactive firefighting to proactive, data-driven management is truly transformative for a global enterprise.

Ultimately, P&G’s experience underscores a vital truth: data quality isn’t just an IT concern; it’s a fundamental business enabler. Investing in robust data quality solutions pays dividends far beyond the technical sphere, directly impacting operational efficiency, risk management, and strategic agility. It’s about turning a tangled web of information into a clear, navigable roadmap for success.

Panasonic’s Data Lineage and Governance: Navigating the IoT Data Deluge

Moving right along, let’s turn our attention to Panasonic’s Smart Mobility Office. In the modern world, companies are swimming in data, especially when they’re at the forefront of connected technologies. Panasonic’s Smart Mobility Office was grappling with massive, massive volumes of information – we’re talking connected IoT device data, real-time weather feeds, intricate geospatial data, and all sorts of other sensor outputs. This isn’t your grandma’s spreadsheet; this is a roaring river of raw data, constantly flowing and accumulating.

The challenge wasn’t just the sheer volume, though that was certainly a factor. It was about understanding where all this data came from, how it transformed, and where it ultimately went. This is what we call data lineage, and without it, managing complex datasets becomes a nightmarish guessing game. Who owns this data? Has it been filtered? Is it reliable? Can I use it for X purpose? These are critical questions for any data-driven operation, especially one dealing with potentially safety-critical applications in smart mobility.

To bring order to this digital chaos, Panasonic deployed Secoda, a centralized data governance platform. Their goal was clear: establish a firm grip on their data, ensuring its quality, security, and usability across various applications. Here’s how Secoda helped them achieve this:

• Tracking and Visualization of Data Flow: Secoda enabled Panasonic to meticulously track and visualize data flow from its origin to its ultimate destination. This capability is like drawing a detailed map of your data’s journey, showing every stop, transformation, and branching path. For a data pipeline involving IoT sensors, cloud processing, and analytical dashboards, this visibility is absolutely invaluable. You can pinpoint exactly where data might be getting corrupted or where a bottleneck exists. It clarifies accountability and makes troubleshooting much, much simpler.
• Role-Based Data Governance: With structured schemas and well-defined roles, Panasonic could establish robust data governance. This means clear rules for who could access what data, who was responsible for its quality, and how it should be used. It moves away from ad-hoc data usage to a more controlled, policy-driven approach. Imagine a system where only authorized personnel can make changes to sensitive mapping data, for instance; this minimizes errors and enhances security, which is super important for smart mobility applications.
• Improved Structured Data Ingestion: The platform streamlined the process of ingesting structured data, making it more efficient and less error-prone. When you’re dealing with live streams of data from thousands of connected vehicles or weather stations, efficient ingestion is non-negotiable. This meant less time wrestling with data formats and more time deriving insights.

Panasonic’s case really highlights the growing importance of data lineage and strong governance, especially in industries that are leveraging the Internet of Things. As more ‘things’ become connected and generate data, the ability to understand, control, and trust that data becomes a competitive differentiator. It’s not enough to collect it; you must govern it, or it becomes a liability rather than an asset. I think this trend will only accelerate, making solutions like Secoda indispensable for forward-thinking organizations.

Virginia Tech’s Data Management Training: Empowering Researchers in the Field

Sometimes, the best technology in the world won’t save you if your people aren’t equipped to use it properly. This brings us to Virginia Tech, where a research group identified a crucial gap: the human element in data quality. In the rough-and-tumble world of field research, data collection can be messy. Think about environmental scientists recording observations in remote locations, or social scientists conducting interviews in various settings. Without proper protocols, training, and a deep understanding of data best practices, even the most meticulous researchers can inadvertently introduce errors or inconsistencies.

This particular group at Virginia Tech recognized that they needed targeted training in data and project management to genuinely enhance data quality in their field research projects. It wasn’t about building a new system necessarily, but about fundamentally changing how their researchers interacted with data from the very first observation. It’s often the small, everyday practices that make the biggest difference, after all.

Their approach was two-pronged and highly effective:

• Emphasizing Data Quality from the Outset: They instilled a culture where data quality wasn’t an afterthought but a core principle from the moment data was conceived, collected, and entered. This involved training on proper data entry, naming conventions, metadata creation, and consistent methodology. It’s about prevention rather than cure. Imagine a new research assistant learning exactly how to record data points in a uniform way, avoiding the ‘quick fixes’ that lead to later headaches.
• Providing Consultative Support: Beyond initial training, they offered ongoing, hands-on consultative support. This meant researchers had a point of contact, an expert to turn to when facing specific data challenges or needing guidance on best practices for their unique projects. This kind of personalized coaching reinforces learning and helps embed good habits deeply within the team’s workflow.

The results were impressive. By focusing on these human-centric interventions, the group achieved substantial improvements in data quality. This wasn’t just about making the data ‘tidier’; it meant their research findings were more reliable, more robust, and ultimately, more impactful. Poor data quality can undermine years of hard work, so getting this right is just incredibly vital. It shows us that investing in people, providing the right training, and fostering an internal motivation for excellence in data management can be just as, if not more, impactful than solely relying on technological solutions. It’s a reminder that even in our hyper-digital world, the human touch remains absolutely critical.

Columbia University’s CAISIS Implementation: Bridging Clinical and Research Data

Healthcare and medical research are fields where data is literally a matter of life and death. Columbia University’s Department of Urology faced a challenge common in many medical institutions: a disconnect between clinical patient care data and research data. Often, these two streams operate in separate silos, leading to redundancy, inefficiencies in data acquisition, and missed opportunities for insight. Imagine a patient’s medical records being separate from the data collected for a clinical trial they’re participating in. This leads to double entry, potential discrepancies, and an overall clunky process.

Their goal was ambitious: integrate research data with patient care data to reduce redundancy and significantly improve data acquisition efficiency. They needed a solution that could speak both ‘clinical’ and ‘research’ languages, if you will. Their answer came in the form of CAISIS (Cancer Archive and Information System), an open-source, web-based data management system. Although originally designed for cancer research, its flexible architecture made it adaptable for broader urology applications.

CAISIS offered several key advantages that addressed Columbia’s specific needs:

• A Scalable Infrastructure: Medical research, especially in clinical settings, generates huge amounts of sensitive data. CAISIS provided a robust, scalable infrastructure capable of handling this volume without compromising performance or security. This meant they wouldn’t outgrow the system quickly, a common pitfall for many data solutions.
• Mirroring Clinical Records: One of the most brilliant aspects was its ability to mirror the representation of clinical records. This wasn’t just about importing data; it was about structuring the research data in a way that aligned with established clinical documentation practices. This congruence minimized friction for clinicians and researchers, making data entry and retrieval intuitive. It’s like having two sides of the same coin, but perfectly aligned.
• Standards-Based Data Exchange: Healthcare data relies heavily on established standards for interoperability. CAISIS supported standards-based data exchange, ensuring that data could be seamlessly shared with other systems, analyzed, and used for various research purposes without cumbersome conversions. This is absolutely critical for collaborative research and for leveraging existing clinical data systems.
• Reduced Redundancy and Improved Efficiency: By integrating these previously disparate data streams, CAISIS dramatically reduced data redundancy. No more double-entering patient demographics or lab results for both clinical care and research studies. This, in turn, led to substantial improvements in data acquisition efficiency, freeing up valuable time for both clinical staff and researchers to focus on patient care and actual discovery.

Columbia University’s implementation of CAISIS offers a powerful lesson in breaking down data silos, particularly in complex, sensitive environments like healthcare. It demonstrates that with the right open-source tools and a thoughtful approach to integration, organizations can create unified data ecosystems that benefit both operational efficiency and scientific advancement. It’s an exemplary case of how smart data management can directly contribute to better patient outcomes and accelerate vital research.

NAV’s Agile Data Management: Embracing the Data Mesh in the Public Sector

Our final case study takes us to the public sector in Norway, specifically to NAV, the Norwegian Labour and Welfare Administration. Public sector organizations are often perceived as slow-moving, bureaucratic behemoths, but NAV’s journey challenges that stereotype. They undertook a significant transformation, shifting from a traditional, centralized data management model to a more distributed, agile approach, commonly known as a ‘data mesh’.

Historically, data management in large organizations often followed a centralized paradigm: a single data warehouse, managed by a specialized data team, acting as the gatekeeper for all data. While seemingly efficient, this model can quickly become a bottleneck, especially for organizations embracing agile software development practices. Agile teams need quick access to data, the ability to iterate rapidly, and often, ownership over the entire lifecycle of their applications, including the data they generate and consume.

NAV recognized this friction. Their centralized data team couldn’t keep up with the demands of numerous agile development teams, each needing specific datasets, different transformations, and rapid deployment cycles. The solution? Adopt a ‘data mesh’ architecture. What does that mean?

• Distributed Data Ownership: Instead of a central team owning all data, responsibility for data shifts to the domain-oriented teams that actually produce and consume the data. For instance, the ‘pension’ team at NAV would own the pension data, ensuring its quality, documentation, and availability as a ‘data product.’ This empowers teams and decentralizes decision-making, accelerating data access.
• Data as a Product: This is a core tenet of data mesh. Data is treated like a product, with well-defined APIs, clear documentation (metadata), and a commitment to quality. Teams that need data can then ‘consume’ these data products directly from the responsible domain team, rather than going through a central IT bottleneck. It’s like having a curated marketplace of internal datasets.
• Self-Serve Data Infrastructure: While ownership is distributed, a common, self-serve data infrastructure underpins the mesh. This provides the tools and platforms (like data lakes, processing engines, governance tools) that domain teams can use to create and manage their data products, without having to build everything from scratch.
• Federated Governance: Instead of a single, monolithic governance body, rules and policies are federated. This means common standards are agreed upon at a higher level, but domain teams have autonomy to implement them in a way that makes sense for their specific data products.

NAV’s transition was not without its challenges, naturally. Implementing a data mesh requires a significant cultural shift, moving from a ‘data as an asset’ mindset to ‘data as a product owned by domain teams.’ It demands new skill sets, changes in organizational structure, and a strong commitment to collaboration. However, the benefits are compelling: increased agility, faster time-to-market for data-driven applications, greater data literacy across the organization, and a more scalable data architecture that can evolve with the business.

This case study brilliantly illustrates the evolution of data management thinking. For organizations striving for true agility and looking to unleash the full potential of their data, the data mesh provides a compelling, albeit complex, blueprint. It’s not for everyone, but for organizations like NAV, it represented a necessary and transformative step forward.

Deeper Dive: Common Challenges in Data Storage & Management

As we’ve seen from these diverse examples, the specific solutions might vary wildly, but the underlying challenges in data storage and management often echo similar themes. Organizations, whether university or corporation, public or private, frequently wrestle with a set of persistent problems. Understanding these is the first step toward crafting an effective strategy yourself.

• Scalability: Data volumes are exploding. It’s not just a trend; it’s the new normal. Every sensor, every transaction, every research experiment generates more data. How do you design a storage solution that can grow effortlessly from gigabytes to terabytes, then to petabytes, without constant overhauls or breaking the bank? Monash University’s LaRDS, with its petascale design, directly addressed this need, providing a future-proof foundation.
• Security and Privacy: In an age of cyber threats and stringent privacy regulations (think GDPR, HIPAA), safeguarding sensitive data is paramount. This isn’t just about preventing breaches; it’s also about granular access control, encryption, and audit trails. How do you ensure only authorized personnel access specific datasets, as LaRDS demonstrated with its permission management, or prevent data leakage like P&G sought to do?
• Accessibility and Usability: Storing data is one thing; making it easily accessible and usable for those who need it is another. Data shouldn’t be trapped in silos or be so poorly organized that it’s impossible to find or interpret. Columbia University’s integration of clinical and research data, and P&G’s efforts to create a ‘single source of truth’, highlight the need for data that is readily available and understandable.
• Data Quality and Integrity: ‘Garbage in, garbage out,’ as the old adage goes. Poor data quality – inaccuracies, inconsistencies, duplications, missing values – undermines everything. It leads to flawed analysis, bad decisions, and wasted resources. P&G’s journey with data quality software and Virginia Tech’s focus on researcher training both underscore the fundamental importance of clean, reliable data. Without trust in your data, what do you really have?
• Data Governance and Lineage: Who owns the data? What are its transformations? How was it collected? Without clear data governance, ownership, and lineage, especially for complex datasets from IoT or multiple sources, the data becomes an untraceable, ungovernable mess. Panasonic’s use of Secoda directly tackled this head-on, offering visibility and control over their data’s entire lifecycle.
• Cost Management: While robust data solutions are critical, they can also be incredibly expensive. Balancing the need for high-performance, scalable, and secure storage with budget constraints is a constant tightrope walk. Organizations need to consider not just initial investment but ongoing maintenance, energy consumption, and staffing costs.
• Integration and Interoperability: Data rarely exists in isolation. It needs to flow between different systems, applications, and departments. Creating seamless integration pathways, whether for research environments like LaRDS, clinical systems like CAISIS, or global ERPs like P&G’s SAP instances, is a recurring architectural challenge.
• Organizational Culture and Training: Perhaps the most overlooked challenge is the human element. Even the most sophisticated systems fail if users aren’t properly trained, if data literacy is low, or if the organizational culture doesn’t value data stewardship. Virginia Tech’s success story is a powerful reminder that people are at the heart of effective data management.

These challenges are interconnected, of course. A failure in one area can quickly cascade into others. But by acknowledging them, and learning from the successes and struggles of others, we can approach our own data strategies with greater clarity and purpose.

Crafting Your Own Robust Data Strategy: A Step-by-Step Guide

Learning from these real-world scenarios isn’t just for intellectual curiosity; it’s about gleaning actionable insights you can apply to your own organization. Crafting a robust data storage and management strategy isn’t a one-time event; it’s an ongoing journey. But where do you even begin? Here’s a structured approach, drawing inspiration from the innovators we’ve discussed:

Step 1: Understand Your Data Landscape – The Grand Audit

Before you build, you must survey the land. Take a comprehensive inventory of your existing data. What types of data do you have (structured, unstructured, semi-structured)? Where is it currently stored? Who owns it? How sensitive is it? What are its lifecycle requirements – does it need to be kept for 5 years, 50 years, or indefinitely? Also, critically, identify your data sources. Are they internal systems, external feeds, IoT devices, or manual inputs? Without this foundational understanding, any solution you implement will be based on assumptions, which can be a costly mistake. Think like Monash or Panasonic, asking ‘What kind of data are we dealing with, and where’s it all coming from?’

Step 2: Define Your Requirements & Goals – What Do You Need to Achieve?

Once you know what data you have, articulate what you need that data to do for you. Are you primarily concerned with long-term archival, like Monash? Do you need real-time analytics for operational efficiency, like Panasonic? Is regulatory compliance and risk reduction your top priority, as it was for P&G? What are your performance requirements (speed of access, processing power)? How critical is data recovery in case of disaster? Setting clear, measurable goals provides a compass for your strategy. Don’t skip this step; it’s where your ‘why’ for data management crystallizes.

Step 3: Evaluate Technologies & Solutions – The Right Tools for the Job

With your requirements in hand, you can now explore the vast landscape of data storage and management technologies. This isn’t just about picking a cloud provider versus on-premise; it’s about considering specific solutions. Are data lakes, data warehouses, or a hybrid approach best? Do you need object storage, block storage, or file storage? What data quality tools, governance platforms, or integration middleware will support your needs? Look at how Columbia implemented CAISIS, an open-source system tailored to their specific integration challenge. Remember, there’s no single ‘best’ solution; there’s only the best solution for your specific context and goals. And don’t be afraid to consider open-source options, they can be incredibly powerful.

Step 4: Prioritize Data Quality & Governance – Building Trust

This step is non-negotiable. As P&G powerfully demonstrated, poor data quality cripples productivity and increases risk. Implement robust data governance frameworks that define ownership, responsibilities, standards, and policies for data creation, usage, and retention. Invest in data quality tools that can identify, cleanse, and prevent errors at the source. Establishing clear data lineage, as Panasonic did, is also crucial, especially in complex environments. Without trust in your data, every analysis, every report, every strategic decision carries inherent risk. Data governance isn’t just a compliance headache; it’s about enabling better, more confident decision-making.

Step 5: Invest in People & Training – The Human Factor

Technology is only as good as the people who use and manage it. Take a page from Virginia Tech’s book: invest heavily in data literacy and management training for all relevant staff, from data entry personnel to senior researchers and executives. Foster a culture where data quality and stewardship are valued and understood across the organization. Provide continuous support and resources. A well-trained workforce that understands the importance of data best practices is your most valuable asset in any data strategy. After all, if the users don’t get it, the system won’t work, will it?

Step 6: Plan for Scalability & Future Growth – A Glimpse into Tomorrow

Your data landscape today will be vastly different tomorrow. Build flexibility and scalability into your chosen architecture. Can it handle anticipated growth in data volume? Can it easily incorporate new data sources or technologies? Think about modular designs, cloud-native solutions, or hybrid approaches that offer elasticity. Monash University planned for petascale capacity from the outset, understanding that research data would only grow. Design for tomorrow, not just for today.

Step 7: Embrace Agility (Where Appropriate) – Adapting to Change

For organizations with complex, rapidly evolving data needs, consider agile data management approaches like the data mesh adopted by NAV. While not for every organization, moving towards distributed ownership, treating data as a product, and fostering self-serve capabilities can significantly increase responsiveness and innovation. This requires a cultural shift, but for dynamic environments, it can unlock tremendous value. It’s about moving away from bottlenecks and empowering those closest to the data.

Key Takeaways and Looking Ahead

These compelling case studies really illustrate that effective data storage solutions are multifaceted, demanding strategic planning, appropriate technology adoption, and continuous investment in both infrastructure and people. Whether it’s implementing a centralized petascale system, enhancing data quality through specialized software, or embracing distributed agile data management, the key, I believe, lies in tailoring the approach to the specific needs, context, and culture of the organization.

We’ve seen how Monash built a digital fortress for research, how P&G tamed a global data beast, and how Panasonic brought clarity to the IoT deluge. Virginia Tech reminded us of the critical human element, Columbia showcased seamless integration, and NAV demonstrated the power of agile, decentralized data. Each story offers a unique lens through which to view the challenges and opportunities of modern data management.

By carefully learning from these diverse examples, and by diligently working through the steps outlined above, researchers and institutions alike can develop robust data storage and management strategies. This ensures not only data integrity, accessibility, and security but also its long-term usability as a powerful engine for innovation and insight. The future is undeniably data-driven, and those who master its management will certainly lead the way.

References

• Monash University’s Large Research Data Store (LaRDS): dcc.ac.uk
• Procter & Gamble’s Data Quality Enhancement: research.aimultiple.com
• Panasonic’s Data Lineage and Governance: research.aimultiple.com
• Virginia Tech’s Data Management Training: datascience.codata.org
• Columbia University’s CAISIS Implementation: pmc.ncbi.nlm.nih.gov
• NAV’s Agile Data Management: arxiv.org