Capgemini has migrated to Proxmox Virtual Environment the infrastructure that processes data from the Sentinel-2B Earth observation satellite, part of the European Copernicus program, for the European Space Agency (ESA). The platform runs on 40 nodes in a hyperconverged design with Ceph distributed storage and two Proxmox Backup Server instances. Proxmox Server Solutions describes it in a success story published on March 4, 2026, which points to per-node licensing cost as the decisive factor behind the choice.
Key points on the Capgemini-ESA project in 30 seconds
- Capgemini, as an ESA partner, moved Sentinel-2B data processing to Proxmox VE.
- The platform uses 40 hyperconverged nodes with Ceph over a 50 Gbps network.
- Backups run on two Proxmox Backup Server instances, incremental, deduplicated, and encrypted.
- Managed Kubernetes in the cloud didn’t meet the connectivity and latency requirements.
- The performance figures come from the success story itself, not independent testing.
Sentinel-2B is part of the Sentinel-2 constellation, which captures multispectral images of the Earth’s surface for uses such as agriculture, forest management, and emergency response. Processing that data stream calls for infrastructure with plenty of fast storage and a stable network. According to the success story, the previous platform was built on legacy clouds, and the plan was to move to a new cloud provider, but Capgemini ended up choosing dedicated infrastructure on physical servers instead.
Why dedicated servers instead of Kubernetes in the cloud
ESA’s requirements were specific: 88 TB of raw capacity, 12 TB of local NVMe storage, and 430,000 random read operations per second (IOPS), plus sustained internal throughput of at least 7 Gbps, expandable to 25 Gbps, and an external API averaging 1 Gbps with peaks of 2 Gbps. According to Proxmox, a managed Kubernetes cloud service couldn’t meet those connectivity and latency constraints, so the solution went with dedicated physical servers, with Kubernetes clusters deployed for pre-production testing.
With 40 nodes, the licensing model weighed heavily on total cost. The success story attributes to Capgemini’s technical management, without naming a specific person, the statement that Proxmox VE “fundamentally changed the economics” of its infrastructure investment, and notes that Proxmox’s subscription model, which charges for support rather than features, turned out to be more transparent. It’s the same argument that has been pushing many companies away from VMware since Broadcom changed its licensing, as we covered in Broadcom’s VMware Price Hikes Are Pushing Customers Toward Proxmox.
The architecture and the migration
The cluster combines compute and storage on the same 40 nodes, connected by a 50 Gbps network, with Ceph integrated into Proxmox VE. According to the success story’s figures, the storage delivers 115 TB of capacity, roughly 7,100 write IOPS and 17,590 read IOPS, and speeds of 3.87 GB/s on writes and 8.89 GB/s on reads. These are vendor-published figures. Backups are split across two Proxmox Backup Server instances, with incremental, deduplicated, and encrypted copies.
The migration was organized into seven phases, from the initial assessment through to the production rollout. Pre-production testing covered backups, recovery, and failover, and workloads only moved to production once they met all the agreed service-level indicators. The success story doesn’t detail how many virtual machines the platform runs or how much the project cost.
For Proxmox, it’s a heavyweight reference: a major European consultancy and an EU space program backing an open-source platform developed in Austria. For anyone weighing a similar move, we’ve covered the key considerations in Migrating from VMware to Proxmox in 2026: A Realistic Guide, Steps, and Hidden Costs.
Frequently asked questions
What has Capgemini done with Proxmox?
It migrated to Proxmox VE the infrastructure that processes Copernicus’s Sentinel-2B satellite data for ESA.
How many nodes does the platform have?
40 hyperconverged nodes with Ceph storage over a 50 Gbps network, and two Proxmox Backup Server instances for backups.
Why did they choose Proxmox?
According to Proxmox, because of the total cost compared with per-node licensing across 40 servers, and because cloud Kubernetes didn’t meet the latency requirements.
Sources:
This article is also available in Spanish: “Capgemini procesa los datos del Sentinel-2B de la ESA sobre 40 nodos Proxmox,” on revistacloud.com.
