Supporting High-Impact Research: SmartX ECP Powers Scientific Platform Development at SIMM, CAS

Jointly developed by the Shanghai Institute of Materia Medica (SIMM) of the Chinese Academy of Sciences and the Shanghai Jiao Tong University School of Medicine, GlycanInsight is an online service platform for the prediction and analysis of carbohydrate-binding pockets, providing high-precision tools for glycan mechanism research and drug discovery. The research was published in Chemical Science, the flagship journal of the Royal Society of Chemistry. As a glycomics algorithmic service platform accessible to the scientific community, GlycanInsight faced typical research-intensive challenges during its deployment, including unpredictable computational demands, frequent developmental iterations, and system maintenance complexities. To address these, the team implemented SmartX ECP, achieving flexible resource management and robust data protection.

Scientific research workloads are often difficult to estimate accurately in advance. During its initial development, GlycanInsight was configured with only 4-core CPU and 16GB RAM. However, under actual operating conditions with concurrent user submissions, the system load surged rapidly, exceeding the capacity of the original resources. SmartX ECP supports online elastic scaling. Following the platform's launch, the development team expanded resources to 64-core CPU and 128GB RAM, enabling the parallel processing of 16 to 36 tasks. This expansion was achieved without manual migration or redeployment, significantly reducing system O&M complexity.

The platform underwent frequent iterations during development, involving joint debugging and third-party data integration by students and non-professional developers, which increased the risk of errors and misoperations. Any improper commit or update error could lead to service disruptions, impacting the user experience. The snapshot feature enables the team to preserve the environment state before critical code or data changes. In the event of a platform anomaly following an update, the rollback function allows for rapid restoration to a stable state, preventing system crashes or prolonged downtime.

To ensure seamless access for international reviewers during manuscript submission, the team required the deployment of two identical service environments across different network links. In a traditional environment, establishing a mirror platform is labor-intensive, involving manual code migration and complex environment debugging. Leveraging the VM cloning feature, the team can rapidly replicate the entire platform—including OS configurations, container services, and essential data—in just a few steps. This ensures consistent deployment across dual platforms, significantly reducing time-to-market while enhancing service accessibility and stability.

Research platforms must expose Web ports for service accessibility while mitigating potential cyberattack risks. In containerized environments, traditional Linux firewalls often fail to effectively restrict port access, as services within containers may bypass host-level firewalls and become exposed to the public internet. Utilizing the network security group features of SmartX (Everoute), the team implemented granular firewall policies at the virtualization layer for the GlycanInsight virtual machines. By enforcing a strict "Allow TCP Port 443 Inbound" rule, all other port access is blocked at the virtual NIC (Network Interface Card) level.

While the core computational tasks of GlycanInsight currently rely on CPUs, the platform plans to integrate AI algorithmic modules and tasks with higher computational densities. This transition will impose new requirements for GPU passthrough and scheduling. SmartX already supports GPU passthrough, with NVIDIA GeForce RTX 4090 (24GB PCIe) GPUs currently deployed and allocated to specific nodes. In the future, GlycanInsight can be seamlessly migrated or expanded to these GPU-enabled nodes without restructuring the underlying architecture, enabling an effortless transition and upgrade from CPU to GPU resources.