The end customer in this project is a leading provider of housing, colocation, and cloud services in Germany. The company operates several high-availability data centers that host customer infrastructures, virtual server instances, and hybrid cloud environments.
Three requirements are paramount in the daily operation of such data centers: high computing density, maximum energy efficiency, and absolute operational reliability. Each server unit must process as many virtual machines or containers as possible simultaneously without compromising performance or stability.
The technological basis is provided by state-of-the-art virtualization clusters that are synchronized across multiple locations. This requires systems that can reliably handle both CPU-intensive workloads such as databases, analytics engines, or container orchestration, as well as fast I/O access to local NVMe storage pools.
Order received: August 2025
Commissioning October 2025
Total volume: CloudDC computing servers in the mid-six-figure range
Table of Contents
Project Description
The goal of the project was to build a new generation of Intel-based CloudDC computing servers to expand an existing virtualization and hosting platform. The systems were to be used within the colocation data center as compute nodes for virtual machines, container clusters, and customer-specific private cloud instances.
The requirements were clearly focused on computing power per height unit, low power consumption, and a robust, scalable architecture. At the same time, the consistent use of NVMe flash memory was intended to achieve significantly higher local I/O performance – for example, for caching, temporary data storage, or VM snapshots within the cluster nodes.
The challenge lay in combining high CPU density, redundant power and network connections, and clear service separation for computing, management, and storage traffic. All systems had to integrate seamlessly into the existing cloud infrastructure and remain expandable in the long term.
Key Requirements
- High computing power per rack unit with optimized power consumption
- Redundant power and network paths for 24/7 operation
- NVMe storage with high IOPS performance for virtual machines
- Support for current virtualization frameworks (Proxmox, VMware, Linux KVM)
- Complete testing and burn-in procedure prior to delivery
- Documented integration in accordance with ISO 9001:2015 / ESD IEC 61340-5-1
Project Implementation
- Server platform
28× Supermicro CloudDC SuperServer (models 621C-TN12R, 121H-TNR, and 511E-WR)
→ Rack-optimized 1U and 2U systems with redundant power supply, efficient cooling, and high packing density. Ideal for virtualization and continuous operation in colocation environments. - Processors (1–2 per server, depending on node)
42× Intel Xeon Gold 6530 (32 cores / 64 threads, 2.10 GHz base, 4.00 GHz turbo, Emerald Rapids, 350 W TDP)
→ 5th generation Intel Xeon Scalable CPUs with PCIe 5.0 and DDR5 support. Offer excellent multi-core performance with high energy efficiency – ideal for virtualized cloud workloads and container platforms. - Memory (140 modules in total)
80× Samsung DDR5-5600 ECC reg. DR (32 GB CL46)
60× Samsung DDR5-5600 ECC reg. DR (64 GB CL46)
→ A total of 6.14 TB DDR5 RAM, distributed across the systems with 512 GB or 768 GB per node. The modules offer maximum stability and data integrity at high clock frequencies. 4 DIMM slots per system remain free for future expansions. - System and data storage
Samsung PM9A3 U.2 NVMe SSDs (1.92 TB / 3.84 TB / 7.68 TB) – a total of 56 drives
→ All-flash storage configuration with up to 6,800 MB/s read and 4,000 MB/s write speeds per drive.
The PM9A3 series offers power loss protection, hardware encryption (AES-XTS), and outstanding IOPS performance for virtualization workloads and VM caching. Operating systems run on separate SATA system drives (Samsung PM893 240 GB). - Network
25 GbE SFP28 connectivity via Mellanox ConnectX-6 LX and ConnectX-4 Lx adapters (dual or quad port)
→ Redundant data and management paths for cluster availability and secure tenant separation. - Assembly & testing
Manufacturing according to ISO 9001:2015 / ESD IEC 61340-5-1
Complete pre-assembly with wired PSUs and pre-installed firmware
Cluster simulation under Proxmox and functional burn-in test (> 24 hours) with I/O benchmarking of NVMe drives
→ Secure, certified delivery as turnkey cloud servers with complete documentation and measurement reports. - Service level
3-year parts warranty (SLA 1/3, 5×9 service) including MUSTANG® systems support
Result
With the newly built server cluster, the customer has a modern cloud infrastructure that significantly increases computing power, storage density, and energy efficiency.
The combination of Intel Xeon Gold 6530 processors (a total of over 1,300 physical and 2,600 logical cores) and 6.14 TB Samsung DDR5-5600 ECC RAM has enabled the virtualization platform to be significantly expanded. Each height unit now offers around 40% higher computing density – while reducing energy consumption per workload.
Local storage has also reached a new level of performance: The 56 Samsung PM9A3 U.2 NVMe drives installed together deliver an aggregate read performance of over 350 GB/s and up to 25 million IOPS. This significantly reduces VM startup times and snapshot processes, and data-intensive container workloads benefit from minimal latencies in the microsecond range.
The redundant 25 GbE connection and clearly separated management network make the platform highly available and easily scalable. New nodes can be integrated without downtime, simplifying future cluster expansions.
The systems were delivered fully preconfigured, tested, and documented. This allowed the operator to put them into productive operation without further customization—a plug-and-play expansion that immediately contributed to the data center's performance improvement.
The result is an energy-efficient, highly scalable cloud platform that meets the growing demands for computing power and I/O performance in colocation operations – and provides a long-term basis for the further expansion of virtual services.