Speakers
Description
The NFDI4BIOIMAGE consortium, operating within the National Research Data Infrastructure (NFDI), is currently developing a comprehensive ecosystem of new file formats, metadata standards, and reusable analysis pipelines. A critical bottleneck in this ecosystem is the physical proximity typically required to handle large-scale, high-resolution image data. Traditional data transfer is often inefficient and can be prohibited by strict security or privacy constraints. To resolve this, we present OSVDI, an Open Source Virtual Desktop Infrastructure (VDI) solution designed to provide remote, interactive, and high-performance access to specialized UI based computing environments. By streaming a familiar desktop interface to the end-user while keeping the data localized, OSVDI extends the reach of imaging facilities and provides a scalable framework for distributed training and collaborative research. This allows researchers to leverage specialized GPUs and optimized workstations from any endpoint—ranging from legacy PCs and thin clients to modern laptops and tablets—without the latency typically associated with remote desktops.
VDI solutions rely on the interplay of various hardware and software components. Unlike commercial vendors who offer these as bundled products, the open-source landscape consists of separate components that must be integrated manually. This often leads to a lack of seamlessness or compatibility issues between modules. OSVDI addresses this by refining existing open-source approaches and developing the missing components needed for a fully integrated solution. Centralized desktop operations are powered by server-based GPUs, providing hardware-accelerated rendering for virtual machines (VMs) and fluid video playback. Using specialized Intel graphics cards, the host hardware is flexibly partitioned into multiple virtual units to support diverse workload requirements. A centralized access gateway provides the entry point for VDI resources. This is integrated into the existing IAM4NFDI variants via Keycloak using OIDC. The technical efficiency of the system is rooted in its ability to maintain high-resolution, graphics-intensive performance while keeping bandwidth requirements moderate, typically requiring only 10 Mbit/s for Full HD and under 100 Mbit/s for higher resolutions. Because the data remains resident near the source at the computing center, data management is significantly simplified and infrastructure security is bolstered.
Beyond individual research, this architecture serves a vital role in distributed online trainings a usecase required by many consortia and research instituions in general. By providing standardized, high-performance training environments, OSVDI ensures that participants in online courses receive a quality of experience comparable to that of local, in-person events.
This development is a collaborative effort hosted by the Computing Center of the University of Freiburg and supported by NFDI4BIOIMAGE and the bwCloud 3 service development project. Committed to the principles of open science, the project provides a publicly accessible demo via DFN-AAI authentication at https://demo.osvdi.uni-freiburg.de/. All source code, interaction pathways, and comprehensive documentation are maintained on GitLab at https://gitlab.uni-freiburg.de/opensourcevdi/, ensuring that the infrastructure can be reused and extended by the broader scientific community.
| background | Potential general purpose service open for various use cases of pretty much all consortia |
|---|