Project background
The client is a Zhejiang-based auto-parts publicly listed company (SRDI). Their existing VMware virtualization licensing was about to expire, and the production-line visual inspection defect-escape rate was as high as 5%. The two goals ran in parallel: localize the infrastructure stack and roll out AI on the production floor.
Client challenges
- VMware perpetual license + SnS renewal cost was high
- 5% defect escape rate — 5–8 customer returns per month caused by missed defects
- High inspector cost — 2 inspectors per line
- Data-localization requirement — preference for domestic hyperconverged
Hongguan's approach
Hongguan delivered a SmartX hyperconverged replacement for VMware (starting at 3 nodes) plus a visual-inspection AI deployment (edge GPU box + open-source vision model). SmartX is compatible with VMware, enabling a smooth migration. The AI model was built on open-source YOLO with a self-trained dataset.
Implementation
- Assessment (1 week): existing VMware environment + workload profiling
- Design (1 week): SmartX 3-node + edge GPU topology
- Migration (2 weeks): P2V → smooth cutover to SmartX
- AI training (4 weeks): 5,000+ defect samples collected, vision model trained
- Rollout (2 weeks): single-line pilot → all 5 production lines
Value delivered
- Resource utilization from 35% to 95% (+60%)
- 35% lower ops cost (SmartX vs. VMware)
- Defect-escape rate from 5% to 0.5% — 5–8 fewer returns per month
- 2 inspectors saved per line, 10 inspectors across 5 lines
- ~¥2M annual combined benefit
Client testimonial
"What surprised us most was that Hongguan didn't just sell products — they asked 'what is the real business problem?' before proposing anything. SmartX and AI are just tools; what mattered was they helped us think through which problem to solve."
— CIO, Zhejiang auto-parts publicly listed company