To meet ESA’s strict performance and latency requirements, Capgemini modernized the Sentinel‑2B processing environment, delivering a dedicated Proxmox VE and Ceph bare‑metal cluster as ESA’s mission‑critical infrastructure implementation partner.
When the European Space Agency (ESA) needed to migrate its critical Copernicus Sentinel-2B satellite data processing infrastructure, the stakes were extraordinarily high. Thanks to Capgemini, Earth observation workloads supporting the Copernicus program demand not just computational power, but a virtualization platform capable of delivering consistent, predictable performance with zero tolerance for downtime.
The original infrastructure, hosted on legacy cloud platforms, could not meet the stringent requirements of the Copernicus Sentinel‑2B production operations. During the migration to a new cloud provider, Capgemini determined that the managed Kubernetes service was unsuitable due to strict connectivity and network‑latency constraints. Rather than accept the limitations of cloud‑managed services, Capgemini chose a different approach: a custom platform built on dedicated bare‑metal servers, with Proxmox VE as the hypervisor foundation. Capgemini’s conclusion was clear: only dedicated infrastructure could consistently deliver the connectivity, latency, and performance characteristics required for mission‑critical Earth observation workloads.
The infrastructure specifications ESA demanded reflect the mission-critical nature of Copernicus Sentinel-2B operations. Internal network connectivity required a minimum sustained throughput of 7 Gbps, with scalability to 25 Gbps for demanding workloads. External API connectivity needed to sustain an average download speed of 1 Gbps with peaks reaching 2 Gbps.
The storage subsystem had to deliver 88 TB of raw capacity complemented by 12 TB of high-performance local NVMe storage, capable of achieving 430,000 IOPS for random read operations. These specifications weren't aspirational targets; they were hard operational requirements. Missing even one of these metrics would degrade data quality and compromise the timeliness of Earth observation products delivered to European stakeholders.
Platform selection criteria extended far beyond performance. The competing solutions offer sophisticated feature sets, but their licensing models created a significant financial barrier when deployed across 40 hypervisor nodes. Total cost of ownership became the decisive factor.
“Proxmox VE fundamentally changed the economics of our infrastructure investment,” noted the technical leadership of Capgemini. “The transparent subscription model and the absence of per‑node licensing constraints allowed us to allocate capital toward performance enhancements rather than software licensing. That strategic advantage proved decisive.”
The implementation deployed 40 Proxmox VE nodes in a hyperconverged architecture, interconnected via a 50 Gbps high-performance network fabric. This topology eliminates single points of failure while providing the elasticity required for variable workload demands across Copernicus data processing operations.
The storage foundation leverages Proxmox's integrated Ceph distributed storage system, delivering 115 TB of usable raw capacity with proven performance characteristics: 7,100 write IOPS, 17,590 read IOPS, 3.87 GBps write bandwidth, and 8.89 GBps read bandwidth. These metrics weren't theoretical; they've been validated through intensive production workloads processing continuous streams of satellite data.
Success required disciplined execution across seven distinct phases:
The pre-production validation phase proved particularly critical. The team deployed Kubernetes clusters in the new environment, migrated core application components, and executed exhaustive backup, recovery, and failover procedures. Only after demonstrating compliance with all service-level KPIs did workloads transition to production infrastructure.
All infrastructure is now protected and restorable via 2 Proxmox Backup Servers, ensuring persistent VM states through efficient incremental backups, deduplication, encryption, and centralized scheduling. The ability to perform fast, granular recovery proves invaluable for a mission-critical system handling irreplaceable Earth observation data.
Today, Proxmox VE powers ESA's Sentinel-2B satellite data processing infrastructure reliably and cost-effectively. The open-source foundation, combined with its robust feature set and transparent economics, continues to deliver the performance and reliability that space-based Earth observation demands.
Raffaele Stefanile
Principal Consultant and Head of Platform Engineering
About European Space Agency
The European Space Agency (ESA) is Europe’s gateway to space. Its mission is to shape the development of Europe’s space capability and ensure that investment in space continues to deliver benefits to the citizens of Europe and the world. Learn more: https://www.esa.int
About Capgemini
Capgemini is an AI-powered global business and technology transformation partner, delivering tangible business value. We imagine the future of organizations and make it real with AI, technology and people. With our strong heritage of nearly 60 years, we are a responsible and diverse group of 420,000 team members in more than 50 countries. We deliver end-to-end services and solutions with our deep industry expertise and strong partner ecosystem, leveraging our capabilities across strategy, technology, design, engineering and business operations. The Group reported 2024 global revenues of €22.1 billion. Learn more: https://www.capgemini.com
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。