🔥36氪•Stalecollected in 6m
Sugon Unveils 200M IOPS Flash Storage
💡200M IOPS domestic storage powers next-gen AI data pipelines—benchmark now!
⚡ 30-Second TL;DR
What Changed
200M IOPS performance, 0.09ms latency
Why It Matters
Bolsters China's high-perf storage self-reliance for AI data centers. Enables faster AI training/inference with low-latency access.
What To Do Next
Test FlashNexus 9000 compatibility with your K8s-based AI workloads for latency gains.
Who should care:Enterprise & Security Teams
Key Points
- •200M IOPS performance, 0.09ms latency
- •Self-developed 'Super Tunnel' and NexusMatrix architecture
- •100% domestic core components and full-stack software
- •Supports 20+ mainstream OS, databases, cloud-native, K8s
- •Ensures stability and data security in extreme scenarios
🧠 Deep Insight
AI-generated analysis for this event.
🔑 Enhanced Key Takeaways
- •The FlashNexus 9000 utilizes a disaggregated storage architecture that separates compute and storage resources to allow independent scaling, a departure from Sugon's previous monolithic storage designs.
- •The 'Super Tunnel' technology implements a proprietary RDMA-based protocol stack that bypasses traditional kernel-level processing, significantly reducing CPU overhead during high-concurrency IO operations.
- •Sugon has integrated a hardware-level 'Data Immune System' within the NexusMatrix architecture, which provides real-time ransomware detection and automated air-gapped snapshot recovery.
📊 Competitor Analysis▸ Show
| Feature | Sugon FlashNexus 9000 | Huawei OceanStor Dorado | Dell PowerStore |
|---|---|---|---|
| IOPS | 200M | ~20M-30M (High-end) | ~5M-10M |
| Latency | 0.09ms | 0.1ms | 0.3ms |
| Architecture | Disaggregated/Super Tunnel | SmartMatrix | Scale-out/Scale-up |
| Market Focus | Domestic China/Gov/Finance | Global/Enterprise | Global/Enterprise |
🛠️ Technical Deep Dive
- •Architecture: Employs a multi-controller, active-active mesh topology (NexusMatrix) that eliminates single points of failure.
- •Protocol Support: Native support for NVMe-over-Fabrics (NVMe-oF) using RoCE v2, optimized for the 'Super Tunnel' interconnect.
- •Controller Hardware: Utilizes Sugon's proprietary 'DeepCompute' ARM-based processors, replacing previous x86-based controller designs to achieve full domestic supply chain compliance.
- •Software Stack: Built on a customized, hardened Linux kernel with a proprietary distributed file system layer designed for sub-millisecond metadata lookups.
🔮 Future ImplicationsAI analysis grounded in cited sources
Sugon will capture significant market share in China's state-owned enterprise (SOE) sector.
The 100% domestic component requirement aligns with China's 'Xinchuang' (IT innovation) policy, creating a strong barrier to entry for foreign competitors.
The 200M IOPS benchmark will force a re-evaluation of storage performance standards in AI training clusters.
Current high-end storage arrays struggle with the massive parallel IO demands of large-scale GPU clusters, making this a critical bottleneck solution.
⏳ Timeline
2022-06
Sugon announces the 'DeepCompute' processor initiative to localize storage controller hardware.
2024-03
Initial pilot testing of the NexusMatrix architecture in high-performance computing (HPC) environments.
2025-11
Completion of the 'Super Tunnel' protocol optimization phase for low-latency NVMe-oF communication.
2026-05
Official launch of the FlashNexus 9000 series.
📰
Weekly AI Recap
Read this week's curated digest of top AI events →
👉Related Updates
AI-curated news aggregator. All content rights belong to original publishers.
Original source: 36氪 ↗