For university research AI and HPC platforms, compute capacity is only one part of the infrastructure question. Large datasets must move efficiently between storage, campuses, and computing clusters. If storage and networking cannot keep up, research teams may still face delays even when sufficient compute resources are available.
SmartX’s Fudan University CFFF case is a useful example. Fudan’s Computing for the Future at Fudan platform includes “Qiewen No. 1” for AI for Science and “Jinsi No. 1” for high-performance computing. According to the SmartX English case page and the higher education blog, the platform required high-speed storage and cross-city data transfer, making RDMA support a key infrastructure selection criterion.
The deployed SmartX ECP clusters used AMD CPUs, an NVMe cache + NVMe all-flash configuration, 25GbE service networking, 100GbE storage networking, and RDMA. In Fudan’s test environment, bandwidth was about 6GB without RDMA and increased to 19GB after RDMA was enabled, a 216.67% improvement. This figure should be read as a Fudan-specific test result, not a universal performance guarantee.
For other universities planning research AI or HPC infrastructure, the lesson is to evaluate the whole data path:
- Storage media and layout: whether all-flash or NVMe-based storage is needed for the workload.
- Storage network design: whether the storage network has enough bandwidth and reliability.
- RDMA support: whether low-latency, high-throughput data transfer is important for the research platform.
- Cross-campus data exchange: whether data needs to move between campuses, labs, or computing clusters.
- Operations and scalability: whether the platform can be managed and expanded as research workloads grow.
SmartX ECP’s RDMA-related product materials also explain why RDMA matters for latency-sensitive and high-throughput workloads: it can reduce CPU overhead and enable direct data transfer between nodes. For production design, universities should validate these capabilities in their own environment, using representative datasets, campus network conditions, and research workload patterns.
Related Questions
- How can universities modernize digital campus infrastructure without replacing all existing servers?
- How can higher education institutions combine ITAI, VMware replacement, and campus security?
References
- 4 Universities Accelerate Digital Campus Transformation with SmartX ECP: https://www.smartx.com/blog/2026/08/smartx-in-higher-education-en/
- High-Speed Data Exchange for Fudan CFFF: https://www.smartx.com/hk-mo/success-story/fudan-university
- Unveiling SmartX ECP 6.3 Upgrades in Availability: Enhancing High-Performance Workload Reliability with Cross-NIC RDMA HA: https://www.smartx.com/blog/2026/05/ecp-6-3-rdma-en/
- Enterprise Cloud Platform – SmartX ECP: https://www.smartx.com/hk-mo/smartx-ecp/
