Azure high-performance computing results in growing wonderful merchandise at Microsoft Floor

This weblog was written in collaboration with the Microsoft Floor and Azure staff. It describes how we used Azure high-performance computing (HPC) to avoid wasting time, prices, and revolutionized our product design piece of producing our Microsoft Floor merchandise.

The Microsoft Floor group exists to create iconic end-to-end experiences throughout {hardware}, software program, and providers that folks love to make use of on daily basis. We consider that merchandise are a mirrored image of the individuals who construct them, and that the correct instruments and infrastructure can complement the expertise and fervour of designers and engineers to ship progressive merchandise. Product stage simulation fashions are routinely utilized in day-to-day choice making on design, reliability, and product options. The group can also be on a multi-year journey to ship differentiated merchandise in a extremely environment friendly method. Microsoft Azure HPC performs an important function in enabling this imaginative and prescient. Under is an account of how we had been in a position to do extra with much less by leveraging the facility of simulation and Azure HPC. 

Floor gadgets improvement on Microsoft Azure 

I’m a Principal Engineer at Microsoft and a structural analyst. I’ve been a heavy person of Azure HPC and an early adopter of Azure A8 and A9 digital machines. In 2015, with the assistance of our Floor IT staff, we deployed and solved many points with Abaqus (a Finite Component Evaluation (FEA) software program) implementation in Azure HPC. By 2016, product stage structural simulations for Floor Professional 4 and the unique Floor laptop computer had absolutely migrated to Azure HPC from on-premises servers. Giant fashions with tens of millions of levels of freedom grew to become routine and simply solved on Azure HPC. This early use of simulations enabled downside fixing for design engineers tasked with robustness and reliability metrics. Utilization grew together with product line progress. Together with my colleagues Pritul Shah, Senior Director of a cross product engineering staff, and Jarkko Sihvonen, Senior Engineer of the IT Infrastructure and Providers staff, we collaborated to scale up structural simulation footprint in our group. The imaginative and prescient to construct a worldwide simulation staff meant entry to computing servers in Western North America and Southeast Asia which was simply deployed by the Floor IT and Azure HPC groups. 

Product improvement: Floor laptop computer  

The supply of Azure HPC for structural simulations utilizing Abaqus helped make this a main improvement instrument for product design. Design ideas created in digital computer-aided design (CAD) programs are translated into FEA mannequin intimately. These are true digital prototypes and represent all main subsystems within the gadget. The analyst can use FEA fashions to impose totally different take a look at and reliability situations in a digital surroundings and decide feasibility. In a couple of days, a whole bunch of simulations are executed to guage varied design concepts and options to make the gadget strong. Subsequently, the chosen design turns into a protype after which topic to rigorous testing for real-world use situations. There are a number of suggestions loops constructed into our engineering course of to match precise exams and FEA outcomes for mannequin validation.  

Within the first graphics depicted above, a digital prototype (FEA mannequin) laptop computer gadget is set-up to drop on its nook to the ground. This fashions the real-world bodily testing that’s performed in our Reliability Engineering labs. The influence velocity for a given top is the preliminary situation for the dynamic simulation. The dynamic drop simulation is executed on a whole bunch of cores of an Azure HPC cluster utilizing Abaqus solver. We used the Abaqus and Express solver which is thought for its strong and correct answer for high-speed, nonlinear, dynamic occasions akin to client electronics drop testing and automotive crashworthiness. These solvers are optimized particularly for Azure HPC clusters and allow scaling to 1000’s of cores for quick throughputs. The simulation jobs full in a matter of some hours on these optimized Azure HPC servers as an alternative of the times it used to take beforehand. The outcomes are reviewed by the analysts and stress ranges are checked in opposition to materials limits. Design groups and analysts then assessment the reviews and make design updates. This cycle continues in very fast loops because the Azure HPC servers allow quick turnaround for evaluations.  

The second graphic depicts an instance of the hinge within the gadget that was optimized for energy. The staff was in a position to visualize the influence induced movement and stress ranges of the hinge inside elements from the simulation. This enabled us to isolate the primary problem and make the correct design enhancements. This perception helped redesign the hinge meeting to trigger decrease stress ranges. Important time was saved within the design course of as just one iteration was wanted for fulfillment. Tooling, bodily prototyping, and testing prices had been additionally saved. 

Presently, your complete Microsoft Floor product line makes use of this strategy of validating design with digital prototypes (FEA fashions) run on Azure HPC clusters. Hundreds of simulation jobs are executed routinely in a matter of weeks to allow cutting-edge designs which have very excessive reliability and buyer satisfaction. 

What’s subsequent 

The staff is now targeted on deploying extra scalable simulation and Azure HPC useful resource for multi-disciplinary groups and for multi-physics modeling. There’s a big alternative to allow machine studying and AI in product creation. Azure HPC and the partnerships inside Microsoft organizations might be leveraged to drive giant scale improvements at a fast velocity. We’re additionally persevering with this digital transformation journey with mannequin primarily based programs engineering (MBSE) with the V4 Institute. World-class organizations trying to do extra with much less and on a quest for scaling digital simulations will drastically profit from collaborating with Azure.  

Be taught Extra 

Azure high-performance computing