Huawei's OceanDisk 1610 now runs ThinkParQ's BeeGFS parallel file system directly on the storage hardware, with 2 PB of raw QLC flash and 175 GBps of bandwidth in a single 2RU chassis. The two companies formalized the work through a strategic co-operation agreement. They will demonstrate the joint appliance at Huawei Connect in Shanghai this week.

The product carries the formal name OceanDisk Built-in File System HPC Storage Solution. It puts the parallel file system inside disk-less storage servers that access a shared pool of NVMe SSDs. Compute and storage are disaggregated. Customers can upgrade compute systems without touching the storage architecture.

What does the hardware actually deliver?

The OceanDisk 1610 is a smart SSD enclosure with a dual-controller active-active design. It provides shared NVMe-over-Fabrics storage pools accessed through RDMA over RoCE. The stated specifications are concrete:

• 36 QLC NVMe SSDs of 161.44 TB each, 2 PB maximum raw capacity
• Up to 175 GBps bandwidth and 5.2 million IOPS from the dual controllers
• 192 processor cores in the compute configuration, from four 48-core CPUs, plus a 1 TB cache

One gap deserves attention. Huawei has not disclosed which CPUs power the enclosure. The report's authors believe they are Huawei/HiSilicon Kunpeng-class ARM server processors. That is an inference, not a confirmed specification. Buyers evaluating the platform should ask Huawei directly.

Why run the file system inside the enclosure?

The integration relies on built-in virtual machine capabilities in the OceanDisk smart disk enclosure. BeeGFS runs on those VMs rather than on separate file system nodes. Zheng Yong, Senior Architect of Huawei Scale-Out Storage, said: "The OceanDisk smart disk enclosure offers built-in VM capabilities to enable partners' file systems to run directly on VMs, delivering a streamlined and efficient HPC storage solution."

The two companies frame the conventional alternative as costly. Traditional HPC storage requires separate deployment of storage devices, file system nodes, and management servers. That architecture is complex, deployment is slow, and operations and maintenance costs climb, they said. The converged design removes dedicated file system servers, shortens deployment, and improves resource utilization, according to both vendors.

These are vendor claims. No third-party benchmark or customer deployment figure accompanies the announcement. The promised simplification is plausible in principle, but the companies have not quantified deployment time savings or utilization gains.

Where does the commercial logic sit?

ThinkParQ frames the deal as a bridge between two markets. CEO Frank Herold said: "We've had a long relationship with Huawei, and we're excited about what this partnership will deliver for customers in both China and Germany. We've recently expanded our operations in China, where we now have a local sales and support team for BeeGFS."

The China connection matters for the sales story. Huawei remains largely excluded from Western procurement on security grounds, so the joint solution realistically targets Chinese HPC and AI installations first, with Germany as the secondary market Herold names. The geographic split is his claim, and no revenue or order figures were disclosed.

The 1610 also supports mainstream third-party file systems including Lustre and GPFS, so BeeGFS gains a certified home on hardware that competitors can also target. That breadth cuts both ways for ThinkParQ.

A note on terminology: Huawei NoF+ is the vendor's branded storage-network stack, meaning NVMe over Fabrics on Ethernet, specifically NVMe over RoCE, plus Huawei additions for lossless transport, failover, and simpler operations.

What comes next?

Huawei Connect in Shanghai this week offers the first public demonstration. Whether the 2 PB, 175 GBps figures hold under real HPC and AI workloads, and whether Huawei confirms the CPU specification, will determine how seriously rival converged-storage vendors need to take the platform.