Over 26,000 researchers supported worldwide
+15% increase in registered users in 15 months
+275% growth in JupyterLab usage
80% reduction in AI model training times
Monthly VRE work sessions grew from 5,340 to 9,904 in 15 months
D4Science adopted a hybrid architecture built using Google Cloud AI optimized Infrastructure to support thousands of scientists worldwide with more scalable and efficient Virtual Research Environments (VREs).
D4Science adopted a hybrid architecture built using Google Cloud AI optimized Infrastructure to support thousands of scientists worldwide with more scalable and efficient Virtual Research Environments (VREs).
From rigid resource management to a scalable scientific infrastructure
D4Science is one of the world’s largest digital science communities, with 217 customized Virtual Research Environments (VREs) and over 26,000 researchers from more than 70 countries. It’s an e-infrastructure managed by Italy’s National Research Council (CNR), historically built on an on-premise setup.
“But in recent years, we’ve had to deal with growing demands for high-performance computing—like training sessions or urgent publications—that went beyond our fixed resource capacity,” says Pasquale Pagano, Executive Director of D4Science.
These limits became especially clear during usage spikes. “For example, a biodiversity research team urgently needed dozens of RStudio instances to meet a tight deadline for a Nature publication. Our local infrastructure was maxed out, causing delays and bottlenecks that could have jeopardized their timeline.”
Shifting to a hybrid model lets us scale flexibly and avoid the delays that used to hold up urgent scientific work.
Pasquale Pagano
Executive Director, D4Science, CNR
To overcome these challenges, D4Science adopted a hybrid model by integrating Google Cloud to boost elasticity and cut provisioning times. “Managed services like GKE, Compute Engine, and Vertex AI gave us scalability, streamlined orchestration, and strong data integration capabilities.”
The hybrid approach allows them to respond quickly to researchers’ needs without disrupting ongoing projects or compromising user experience. “Shifting to a hybrid model lets us scale flexibly and avoid the delays that used to block time-sensitive scientific work.”
Cloud-native AI for next-generation research
Using GKE and Compute Engine, D4Science deploys key services like JupyterLab, RStudio, Galaxy, and WebODV in containers that are dynamically orchestrated and scaled based on demand with the Google Cloud AI optimized infrastructure. These services integrate with on-premise storage and core systems, ensuring continuity and data control.
Vertex AI lets us build AI tools that put users first and make working with data easier and smarter.
Pasquale Pagano
Executive Director, D4Science, CNR
Vertex AI was introduced to build AI tools such as semantic search for data catalogs and conversational interfaces that simplify scientists’ access to APIs. “The goal is to democratize AI, making it accessible even to those without programming skills,” Pagano explains. Early results show up to an 80% reduction in AI model training times. “Vertex AI lets us design user-centered AI tools that make interacting with data smarter and easier.”
Google Cloud’s robust AI/ML toolkit, data interoperability, and commitment to open-source standards were key to making this possible. D4Science is also exploring additional services like BigQuery, Dataflow, and managed databases on Google Cloud to enhance analytics and research support.
Flexible VREs that empower researchers
With 217 active VREs and over 26,000 users across more than 70 countries, D4Science serves as a digital enabler of global science. Google Cloud services support dynamically allocated computing environments, while core technologies—especially foundational ones—and data catalogs remain on-premise to ensure control and compliance.
This hybrid model has proven effective: monthly VRE work sessions jumped from 5,340 to 9,904 in just 15 months. JupyterLab usage rose by 275%, peaking at 1,217 sessions, while RStudio usage increased by 37%, reaching 592 sessions. Registered users also grew by 15% over the same period. These gains reflect improved accessibility, responsiveness, and user satisfaction—driven by the Google Cloud-powered architecture.
Sustainability, hybrid resilience, and regional data governance
D4Science’s commitment to data privacy and sovereignty is central to its architecture. Sensitive data remains on encrypted local storage, while public or anonymized datasets can be processed in the cloud.
“Google Cloud’s regional infrastructure in Europe, strong GDPR compliance, and industry-leading sustainability practices align with D4Science’s mission,” Pagano confirms. The ability to choose data residency, securely integrate with existing systems, and benefit from carbon-neutral operations enhances both compliance and reputation. “Positioning Google Cloud as a sustainable technology partner also strengthens D4Science’s participation in funding calls.”
By keeping catalog systems and persistent storage on-premise while enabling scalable cloud workloads, D4Science strikes an optimal balance of control, flexibility, and performance.
Positioning Google Cloud as a sustainable technology partner also supports D4Science’s participation in funding calls.
Pasquale Pagano
Executive Director, D4Science, CNR
D4Science is one of the world’s largest digital science communities, with 217 customized Virtual Research Environments (VREs) and over 26,000 researchers from more than 70 countries. It’s an e-infrastructure managed by Italy’s National Research Council (CNR), historically built on an on-premise setup.
Industry: Public Sector
Location: Italy
Products: Google Cloud, GKE, Compute Engine, Vertex AI, Cloud Storage, Workspace