Upstream Applied Science accelerates innovation with HPCBOX on Azure to reduce turnaround time for modeling methodology assessment

Accelerating innovation with high-performance computing on Azure

Simon Hubbard founded Upstream Applied Science Ltd. in the United Kingdom in 2019 to offer applied science and math consultancy services and specialized CFD capabilities. The company develops and applies open source toolsets in the computer-aided engineering space and employs cloud computing capability to enable the use of more advanced computational techniques in product design and development. 

Upstream contacted Drizti to help it assess the performance of CFD models. Drizti’s HPCBOX solution on Microsoft Azure delivers a desktop-centric, workflow-enabled HPC platform with a rich user experience. It lets users harness the Azure big compute infrastructure and run applications in parallel with their native GUI to significantly reduce the time to develop products and allow users to focus on innovation.

“Microsoft Azure’s capabilities are a perfect fit for HPC workloads. We are very excited to work with innovative companies like Upstream by partnering with Microsoft and delivering our HPCBOX platform through Azure Marketplace,” said Devarajan Subramanian, Chief Technology Officer, Drizti.

Determining the best modeling approach with HPCBOX

For turbulent flows, the choice of how to represent the turbulence in a CFD model is critical to the validity of the results. There are many low fidelity models of turbulence that model its effects entirely. These Reynolds’ Averaged Navier Stokes (RANS) models are often referred to as engineering workhorse models as they are relatively easy to use and are computationally efficient. With the advent of cloud computing, however, higher fidelity Large Eddy Simulation (LES) turbulence models that model only around 10% of the effects of turbulence and capture the remainder directly can now be brought to bear on CFD studies. 

Upstream wanted to assess the performance of a number of low fidelity and one high fidelity turbulence model on the Food and Drug Administration's (FDA) CFD benchmark model of a blood pump to determine which modeling approach was beneficial ahead of a CFD phase of product design and development.

Using HPCBOX on Azure, Upstream was able to assess several model configurations in addition to the desired variations in turbulence modeling approach. HPCBOX allowed Upstream to run a range of CFD models of the blood pump and retrieve data for analysis quickly and effectively. 

HPCBOX on Azure delivers the tools for success

“HPCBOX helped in addressing the challenge by providing quick and straightforward access to Microsoft’s Azure HPC capability, offering a toolset that enabled quick setup and execution of computational workflows,” Hubbard said.

Upstream completed its turbulence model simulations in less than five days, showing the turnaround time that can be achieved for performing modeling methodology assessment ahead of a computational phase of product design and development.

“HPCBOX is a great tool for facilitating fast and effective access to the Microsoft Azure HPC environment,” Hubbard said. “The support provided by Drizti during the trial was very responsive and proactive in releasing performance-enhancing developments to the toolset.”