Solution Architecture

To meet HPC needs of various CAD/CAE applications, Sugon recommends the following solution.

Design system:The design system simultaneously meets design/simulation users’ design and pre-/post-processing needs. Sugon’s design-system solution can provide Sugon dedicated high-end graphics servers with high per-core performance, large per-core memory, good memory performance, 3D visualization, and high local I/O, thereby meeting urgent multi-user remote pre-/post-processing and design needs. Graphics servers can be remotely centrally deployed, with unified management, diverse services, and data-security assurance.

Simulation system:To meet good parallel scalability of CFD-class application software, provide Sugon flagship TC4600E blade products to raise compute density and strengthen computing efficiency, meeting parallel-computing needs of cases above 100-million-class meshes. For electromagnetic-simulation and structural-analysis (implicit solving) simulation software with weak cross-node parallel scalability and high memory-capacity/performance needs, provide Sugon TC6600 4-socket blades, I840 4-socket servers, I980 high-end 8-socket servers, and more, flexibly combined to meet diverse application needs. In addition, X580/745/780/785/795/ and other GPU-server products can be provided for GPU-accelerated application software. Formulate professional, personalized simulation schemes by user application characteristics and case scale. The simulation system can seamlessly integrate with the design system, forming a mature integrated solution.

Storage system:Most CAE simulation software has good parallel scalability and large case data; to handle massive data from high-throughput computing, Sugon high-performance clusters are configured with ParaStor distributed parallel storage. ParaStor supports massive data and online expansion, with storage capacity and aggregate bandwidth rising together, meeting high-throughput computing’s needs for the storage system.

Network system:Sugon equips high-performance clusters with high-bandwidth, low-latency InfiniBand or Omni-Path and other high-speed networks, meeting data-exchange needs of various MPI programs in application software, raising multi-node parallel efficiency, and speeding simulation.

Software system:Sugon Gridview cluster-management and job-scheduling system provides an industrial-simulation cloud edition meeting manufacturing-industry requirements, realizing a unified portal for design and simulation platforms, integrated management, monitoring management and job scheduling, and graphical portals of commonly used mainstream commercial CAE software matching user habits, supporting portal custom development, and providing an HPC base runtime.

O&M services:Sugon EasyOP online O&M is 7×24 online O&M for high-performance clusters, ensuring good cluster runtime, freeing researchers from tedious cluster-management work so they can put more time into actual research.