Fastone achieves new high performance computing levels with Oracle Cloud

High Performance Computing (HPC) is the foundation of computer-aided engineering, simulation, and machine learning in many industries, including life sciences, semiconductors, and automobile manufacturing. With its powerful, cost-effective computing capabilities to solve complex mathematical and scientific problems, HPC is also a necessity in higher learning and research sectors.

Shanghai-based Fastone Information Technology provides a one-stop, application-defined cloud platform optimized for over 240 HPC applications, applicable to numerous use cases when deployed on-premises, in the cloud, or in hybrid scenarios. Using these services, the company’s customers experience increased efficiency and reduced total cost of ownership.

Fastone’s leadership team recognized the need to add massive amounts of public cloud computing resources to support its customers’ emerging requirements for scaling to demand. The goal was to build an HPC infrastructure that could automatically scale to more than 10,000 cores.

The company invited cloud platform providers, including Oracle, to present a benchmarking proof of concept for HPC workloads. The challenge was to perform a LINPACK benchmark, documenting cloud efficiency at a level that could vault Fastone into the TOP 500 rankings.

Oracle excelled in this test of computing titans, with Oracle Cloud Infrastructure (OCI) in an HPC shape using Oracle Optimized3 instances running at a very high bandwidth and low latency made possible by Remote Direct Memory Access. Using Intel Ice Lake-based processors, the proof of concept scaled to 270 nodes/9200 cores with RoCEV2 (RDMA over converged Ethernet).

As a result, Fastone selected OCI for superior HPC performance, driven by the latest generation of Intel Ice Lake processors, and for the unmatched ability to scale with stability while also providing high volume storage.

OCI provided Fastone with the capabilities to offer HPC workloads at a lower cost, which enabled the company to extend this benefit to its cost-sensitive customers. The company’s state-of-the-art CPU, GPU, and HPC-related technologies speed up computer-aided engineering processes. And these ready-to-use resources greatly reduced the time from procurement to simulation for Fastone clients.

By offering flexible computing services on bare metal servers optimized for HPC applications, Fastone’s customers can build and schedule massive cloud resources in a short period of time. This platform compressed the work of years into only hours for customers requiring HPC for drug discovery, or weeks into hours for electronic design automation, computational fluid dynamics, and many others.

OCI autoscaling allowed Fastone to run an HSPICE analog circuit simulation 42 times faster by dynamically turning on the required computing resources. This provided greater chip performance for critical applications, such as vehicle braking systems, robotic surgical equipment, or 24/7 manufacturing lines. By automatically monitoring the number of tasks and resources, customers reduced simulation time from 30 days to 17 hours, while silicon chip designs have grown larger and more complex.

Likewise, in drug discovery, the HPC on OCI solution equipped one customer  with a 2,875X increase in AutoDock Vina molecular modeling simulation capabilities, the docking analysis system governing the identification of new medicines. In another example, one customer needed to predict how small molecules bind to a receptor of a known 3D structure. Fastone reduced simulation time for 28 million molecules from 5 years to 15.23 hours by automatically building and scheduling large-scale computing power clusters with 100,000 cores on OCI.