OpenNebula 7.2 adds gRPC API for sovereign AI clouds

OpenNebula Systems has launched OpenNebula 7.2, an update aimed at sovereign cloud and AI infrastructure deployments.

The release adds a new gRPC-based API to improve communication between platform components in environments with large numbers of concurrent operations. The change is intended to reduce latency and improve responsiveness as AI and high-performance computing workloads move into production.

It also introduces real-time virtual machine execution logs in the Sunstone web interface, giving administrators direct visibility into workload behaviour without relying on command-line tools.

AI Focus

A central part of the update is tighter integration with NVIDIA hardware and networking products used in GPU-heavy computing environments. OpenNebula 7.2 integrates NVIDIA Fabric Manager to orchestrate NVSwitch and NVLink interconnects in multi-GPU systems for AI training and high-performance computing.

The platform has also been validated with NVIDIA Spectrum-X networking platforms, NVIDIA Grace Blackwell GB200 systems and NVIDIA BlueField DPUs. Those validations cover support for low-latency Ethernet fabrics, VSwitch topologies and network offloading, as well as hardware-level isolation and tenant segmentation.

The release includes Enhanced VM Compatibility, or EVC, to make it easier to move workloads across different hardware environments. That is intended to reduce disruption during hardware refreshes and platform upgrades.

Networking updates also include IP address sharing for groups of virtual machines, a feature designed to simplify deployment of multi-tier applications. Production-ready LXC drivers have also been added, bringing lifecycle management features such as network interface hot-plugging and disk snapshots to Linux containers.

Security Changes

OpenNebula 7.2 expands confidential computing support for KVM workloads through hardware-rooted trust and memory encryption. The update also adds virtual TPM integration and gives administrators the option to enforce mandatory two-factor authentication for all Sunstone users.

These additions are aimed at regulated and sovereign cloud environments, where operators face growing pressure to demonstrate stricter controls over workload integrity, data protection and user access.

Storage Updates

Storage is another major area of change in the release. Storage virtual machines can now be live-migrated between LVM and file-based datastores, including shared, local and thin-provisioned backends, without downtime.

This should allow maintenance and storage rebalancing with less disruption. The platform also now integrates natively with Everpure, formerly Pure Storage, FlashArray through the FlashArray REST API for block storage lifecycle management.

Further storage changes include multi-tier caching for local datastore drivers and tuning for LVM backends to reduce input and output bottlenecks. Backup support has been extended with incremental backups for NetApp systems, along with updates to Veeam and oVirt API integrations intended to improve restoration reliability.

Deployment Support

OpenNebula has also introduced OneForm, an automated service for on-demand deployment and configuration of clusters in distributed environments. OpenNebula 7.2 also adds support for SUSE Linux Enterprise 15 and the latest versions of AlmaLinux 10, RHEL 10, OpenSUSE 16 and Debian 13.

The update comes as infrastructure software suppliers position themselves around demand for AI systems that can run in private and hybrid environments rather than only on public cloud platforms. They are also responding to growing interest from governments, regulated industries and large enterprises in systems that keep operational control and data handling closer to home.

OpenNebula develops cloud management and virtualisation software for private, hybrid and edge infrastructure. It says it has seen rising adoption as organisations look for alternatives to VMware while trying to preserve existing hardware and operating practices.

The company says its software is used in more than 5,000 cloud deployments worldwide, including infrastructure that scales to thousands of hosts and tens of thousands of GPUs.

"Organisations deploying AI today need infrastructure that is not only powerful but also easy to operate and secure by design. OpenNebula 7.2 enables our users to build and manage AI Factories and sovereign cloud environments with the confidence that their infrastructure can scale to meet the growing demands of production AI," said Dr. Ignacio M. Llorente, Managing Director, OpenNebula Systems.