Technology Partners

The Supermicro SuperServer comprises two independent server nodes sharing 24 NVMe PCIe Gen5 drives in a 2U form factor. This Storage Bridge Bay (SBB) server is ideal for implementing high-availability NVMe drives.

Each node supports:

1. Single Socket E (LGA-4677) 5th/4th Generation Intel® Xeon® Scalable processor, up to 350W TDP
2. 2 PCIe 5.0 x16 HHHL slots
3. 2 PCIe 5.0 x8 HHHL slots
4. 8 DIMMs per node with 1DPC, up to 2TB memory capacity with 8 DIMMs of 256GB 3DS RDIMM DDR5-5600 ECC memory per node
5. Dedicated 1GbE Private Ethernet connection for node-to-node communication (Heartbeat)
6. Redundant Titanium 2000W Power Supplies

The system was equipped with 24 KIOXIA CM7-R KCMYXRUG1T92 drives. To fully utilize the drive PCIe bus connections, each drive was split into two namespaces. The first namespaces were used for RAIDs active on Node0, while the second namespaces were used for RAIDs active on Node1. To fully utilize the drive performance, we created a symmetrical configuration of relatively small RAIDs to be close to most real-life configurations. As a result, 3 RAIDs of 8 namespaces each were created at each node. The following picture and table showcase the namespace mapping.

The “strip size” (also known as “chunk size”) is the minimum amount of data that is written on one single drive within the RAID.

In case of failure of one of the two nodes, the RAID groups on the failed node are automatically migrated to the surviving one, as explained in the image below.

The following table reports the results obtained during simultaneous testing of both nodes' RAID groups or NVMes. Tests on RAID groups were conducted with 32 threads/64 queue depth per RAID group.

Expected RAID performance is based on RAID architecture:

• Sequential read performance should approximate the combined sequential read performance of all the drives within the RAID groups.
• Sequential write performance should approximate the combined sequential write performance of all drives minus the parity drives.
• Random read performance should match the combined random read performance of all the drives.
• Random write performance in RAID 6 should be around 1/4 of the combined drives' random write performance, depending on the drives' random read/write performance ratio and CPU resources.

Read the full whitepaper

Today, applications demand not only exceptional performance but also uncompromising resiliency. The combination of Supermicro SuperServer SSG-221E-DN2R24R and Xinnor's xiRAID Classic creates a highly efficient and robust Cluster-in-a-Box solution. This architecture provides unprecedented performance, utilizing the full potential of NVMe PCIe Gen5 drives while ensuring high availability and seamless failover capabilities.

The solution achieves remarkable efficiency, consistently operating at 95% to 100% of the raw hardware performance across various workloads. This optimization level ensures organizations maximize their investment in cutting-edge storage technologies while maintaining the reliability required for mission-critical operations. With resilience to multiple drive failures and a complete server node failure, the platform delivers uninterrupted performance and exceptional data protection.

By combining innovative hardware and software technologies, this joint solution addresses the evolving needs of modern storage systems, setting a new standard for performance, reliability, and scalability.

As a global leader in high performance, high efficiency server technology and innovation, Supermicro develops and provides end-to-end green computing solutions to the data center, cloud computing, enterprise IT, big data, HPC, and embedded markets. Their Building Block Solutions approach allows them to provide a broad range of SKUs, and enables to build and deliver application-optimized solutions based upon customers’ requirements.