Technology
Compute-storage disaggregation + EBOF all-flash. The only team in China achieving compute acceleration via hardware-accelerated all-flash.
Compute-storage disaggregation
- NVMe-oF: physical decoupling of compute and storage
- RoCEv2 high-speed interconnect
- IPv6 dynamic addressing, node identity and address unified
- SDN intelligent scheduling
Latency ≤100μs · Bandwidth ≥100Gbps · Expansion ≤4h
EBOF all-flash storage
- PCIe Gen5 gateway-free distributed storage pool
- 8+2 EC erasure coding, 20% redundancy
- Smart migration, lifespan +20%
IOPS ≥1M · Bandwidth ≥56GB/s · 99.9999% durability
Huawei benchmark (same cost)
| Metric | Improvement |
|---|---|
| Data load time | -72% |
| Throughput | +64% |
| 3Y TCO | -30%+ |