Architecture
MarghERita is the Big Data Platform of the Emilia-Romagna Region, designed to support Public Administrations in managing, processing and enhancing large volumes of data. The platform provides an integrated ecosystem of services for data engineering, data science, artificial intelligence, business intelligence and advanced application development, enabling innovative projects in the fields of healthcare, environmental protection, mobility, road safety and territorial governance.
Designed according to the principles of openness, modularity and portability, MarghERita is mainly based on open source technologies and can evolve over time by integrating new tools and new analytical paradigms without proprietary technology constraints.
Technological infrastructure
The infrastructure is distributed across the regional data centers in Bologna and Ferrara and represents one of the main computing environments dedicated to data analysis for the regional Public Administration. The platform includes:
- 75 computational nodes;
- Intel Xeon Gold 6336Y processors;
- NVIDIA Tesla T4 GPUs;
- 512 GB of RAM per node;
- approximately 1 PB of storage capacity distributed on DELL Isilon technology;
- a virtualized infrastructure based on Kubernetes, Hadoop and KVM.
This architecture ensures high processing capacity, scalability and reliability, enabling the simultaneous execution of multiple projects and workloads. Resource virtualization and the use of dedicated environments also allow logical segregation of data and applications among the different projects hosted on the platform.
From Big Data to Artificial Intelligence
MarghERita continues its technological evolution to support increasingly advanced analytics and artificial intelligence scenarios.
In addition to the resources already available, a new specialized infrastructure composed of 8 NVIDIA L40S GPUs is being introduced, designed to support machine learning, deep learning and Generative Artificial Intelligence applications. These new resources expand the platform's capacity to train and run large language models (LLMs), intelligent agents, Retrieval Augmented Generation (RAG) systems and AI solutions supporting public decision-making.
The infrastructure enhancement also represents a strategic element for strengthening the sovereignty of public data and enabling Public Administrations in the region to safely experiment with new Artificial Intelligence applications within an environment that is governed, controlled and managed directly by the regional ecosystem, while maintaining full control over data, models and computational resources.
MarghERita is thus evolving from a platform dedicated to Big Data into an infrastructure capable of supporting the entire lifecycle of Artificial Intelligence solutions, from experimentation to production deployment.
Software architecture
The platform adopts a Data Lakehouse model based on open source technologies integrated into a unified ecosystem for data management, processing and enhancement. The technology stack includes:
- Kubernetes for orchestrating containerized services;
- Rancher for managing Kubernetes clusters;
- Hadoop for distributed processing;
- HDFS as a distributed file system;
- Apache Hive for data warehousing;
- Apache Iceberg for advanced dataset management;
- KVM for hardware resource virtualization;
- OpenNebula as a component for managing the virtualized infrastructure.
Service catalogue
Data Engineering
Services dedicated to data acquisition, transformation and orchestration:
- Apache NiFi·
- Apache Airflow;
- Apache Kafka;
- Apache Spark;
- Trino.
Data Science and AI
Tools for advanced analytics and algorithm development:
- JupyterHub;
- Python and R libraries;
- frameworks for machine learning and artificial intelligence;
- integration with Apache Spark;
- OpenWeb UI for creating chatbots and agents based on language models.
Business Intelligence
Services for data visualization and sharing results:
- ·Apache Superset for dashboards and interactive analytics.
DevOps and Deployment
Tools dedicated to application release and management:
- · GitLab;
- Harbor;
- ArgoCD;
- Argo Workflows;
- FileBrowser;
- sFTP.
Security, governance and resource isolation
The platform is designed to ensure high levels of security, access control and data governance.
Identity and user profile management is handled through Keycloak, while data access control is governed by Apache Ranger. Infrastructure monitoring and cluster administration are supported by Rancher, Apache Ambari and centralized logging, audit, backup and recovery systems.
The platform's multi-tenant architecture makes it possible to create dedicated working environments for individual projects through Kubernetes namespaces, databases, data areas and segregated application services. This approach ensures isolation of computational resources and data among different projects, in line with the security, confidentiality and governance requirements of Public Administration contexts.
Usage models
MarghERita provides three usage models.
Fully Managed
Users work with tools and services centrally administered by the MarghERita team, focusing exclusively on developing solutions and analyses.
Self Managed
Project teams can develop and deploy customized applications on Kubernetes while maintaining control over the software and release lifecycle.
Hybrid
This model combines centrally managed services and customized applications, making it possible to integrate data pipelines, analytical models and tailor-made services within a single technological ecosystem.

Request information
Organizations interested in learning more about the platform's features, proposing new projects or requesting technical support can contact the MarghERita team through the dedicated channels available on the institutional portal. The platform management team provides support in defining project architectures, ingesting data, configuring working environments and adopting best practices for security, privacy and data governance.
For assistance and clarification requests, the dedicated mailbox margherita@regione.emilia-romagna.it is available.
