Sugon Storage System Loses Top Spot Due to Transparency Rules

Sugon’s ParaStor F9000 all-flash storage system has lost its top position on the IO500 Production list. The benchmark committee reclassified the hardware to the Research list because it failed to meet strict reproducibility standards. This decision restores Intel-backed Aurora to the lead in the main category, emphasizing that high performance alone is not enough to hold the top spot if the underlying technology is opaque.

Reproducibility standards define the top tier

The IO500 benchmark separates its rankings into Production and Research categories. The Production list requires that the system’s architecture is well-documented and the software is generally available. This allows independent researchers to understand and potentially replicate the results. Sugon’s submission lacked these widely available details, forcing the committee to move it to the Research tier, which accepts more experimental or proprietary configurations.

According to Tom's Hardware, this distinction is critical for the credibility of the rankings. While the Sugon system achieved impressive raw numbers, its proprietary nature meant the results could not be independently verified in the same way as open-source alternatives. The committee explicitly cited the lack of architectural transparency and limited general availability of the file system as the reasons for the demotion.

Aurora retakes the lead in production

With the Sugon systems removed, Aurora from Argonne National Laboratory has reclaimed the top spot on the Production list. Aurora uses a custom storage subsystem featuring Intel Optane Persistent Memory and the DAOS file system. Unlike the Sugon setup, DAOS is open source and extensively documented. This allows the community to examine the architecture and understand how the high performance was achieved, satisfying the benchmark's transparency requirements.

The performance gap was significant before the reclassification. Sugon’s submission scored roughly 2.5 times higher than Aurora in the main category. However, because Aurora meets the higher standards for reproducibility, it remains the benchmark holder for production-grade systems. This highlights a practical trade-off: buyers seeking the absolute highest speed may look at proprietary options, but those needing verified, reproducible performance often favor open architectures.

Transparency remains a key trade-off

This decision underscores a broader trend in high-performance computing. While bespoke, proprietary systems can push raw metrics higher, they often lack the openness required for top-tier recognition. For organizations evaluating storage systems, this serves as a reminder that raw speed is only one part of the equation. The ability to audit, understand, and reproduce performance is a critical factor, especially for scientific and industrial applications where reliability and verifiability are paramount.