AWS Partner Story: OnScale

Cars. Electronic devices. Motors. Planes. Roads. Space exploration. Structures. All built by engineers. And our world evolves through the continuous innovation of engineers. But what happens when engineers can’t access the compute power and software resources that they need to experiment, simulate, and pursue new breakthroughs in their respective fields?

The world may keep spinning, but a lot of the innovation that powers our world slows down. And a lot of problems are left unsolved. The biggest bottlenecks keeping many engineers from breakthrough designs and industry disruption are traditional licensing models and hardware costs, which can prevent startups and global corporations alike to run simulations quickly and at optimal scale.

With its new Solver-as-a-Service platform, OnScale hopes to change the way engineers across disciplines access the resources they need.

“We want to bring about the future of engineering,” says Ian Campbell, CEO at OnScale. “Our mission is to give engineers seamless access to high-end computing engineering software, multi-physics solvers, and a cloud HPC computing platform where they can run extremely large analyses using AWS to get results back quickly and cost-effectively.”

“The more power you can give an engineer, the more complex problems they’ll solve,” says Gerry Harvey, vice president of engineering at OnScale.

“At a previous company the main blocker for product development was our engineering tools and engineering software,” says Campbell. “We felt we were always sitting around waiting for data to come back to give us engineering insights. Inevitably we’d have to pull the trigger and take a new design to production, and it was very painful when we’d do so and then realize that it didn’t work as intended. If we’d been able to conduct better engineering analyses at the front-end we could have solved most of these errors through virtual prototyping while spending less time and money on physical prototyping.”

Prior to moving to AWS, OnScale ran similarly to other CAE companies. “We had a fixed license for a given computer or a floating license to use on a network. Basically, the power of the solution was limited to the hardware that you had available to you,” says Robbie Banks, vice president of product at OnScale.

The team sought to address these pain points by taking the company’s existing technology and building a cloud-first product using OnScale’s multi-domain, multi-physics solvers, taking advantage of the nearly limitless scalability and availability of AWS, and providing software as a service (SaaS) pricing options for customers. “We’d been listening closely to our customers’ feedback, and what we found was our customers wanted to move faster, run more powerful simulations, and run simulations more often,” explains Harvey. “When we discussed with clients the option to run CAE jobs in the cloud and take advantage of SaaS subscription pricing, the positive response was overwhelming.”

Soon, the concept for OnScale’s Solver-as-a-Service platform was born. But while the team fostered a unique and clear vision of the solution they wanted to provide customers, they needed help executing their vision of taking their current infrastructure and migrating to AWS.

For re-architecting, migration, and implementation expertise on AWS, OnScale turned to the experts at nClouds.

OnScale sought to identify an accomplished AWS Partner who could help them meet aggressive timelines to migrate to AWS and architect a solution optimized to take advantage of AWS capabilities. “We have expertise on the solver side and in the engineering-specific requirements,” says Banks. “But we were trying to get to grips with AWS technology. Our AWS account manager told us there was a stable of AWS Partners who were well-recognized and accredited by AWS, and we did a bit of due diligence, and we chose nClouds.”

nClouds strives to differentiate itself through a unique customer engagement model by embedding within a customer’s team, coupled with deep skills. “We partner with our customers, working as extensions of their teams to build and manage their infrastructure,” explains JT Giri, CEO at nClouds. With an ever-growing list of AWS certifications, including AWS Advanced Consulting Partner, AWS Managed Service Partner, AWS Approved Well-Architected Partner, AWS DevOps Competency Partner, AWS Standard Technology Partner, and AWS Channel Reseller Partner, the nClouds team has gone all-in with AWS and makes it a goal to demonstrate its deep AWS expertise to customers from the first time they connect onward. “Building and demonstrating AWS credentials is important to us,” says Giri. “We see a persistent gap in the marketplace for the combination of deep technical skills and experience that we bring across the breadth of DevOps and modern infrastructure solutions on AWS. But it’s the collaboration with clients who bring their business and domain expertise that results in real innovation.”

OnScale established a fully integrated workflow process with nClouds, and the team got to work migrating OnScale to AWS and architecting a sophisticated solution on AWS taking advantage of native managed services wherever possible. “Within 6 – 8 weeks of development, we had a Beta out to some of our customers,” says Campbell. “nClouds helped us meet and exceed the aggressive timelines we’d set.”

nClouds' focus was to migrate OnScale's Solver to AWS. OnScale provided the various executables for different types of simulations that can run. nClouds reviewed the executables and determined that the best path would be to build all those capabilities into a Docker container that could easily execute in the cloud. Thus, the same solver that clients were running on their laptops could be used on AWS cloud instances, but with massively more compute power. After the container was built and tested, it was a matter of linking OnScale's desktop application to the newly created cloud capabilities.

OnScale’s SaaS solution takes advantage of AWS to run complex simulations in parallel securely and quickly. “Our software itself is C++ and various other stacks. We need to be able to take the model that we have, deploy it to AWS, and solve it using AWS resources,” explains Banks. To meet and exceed its scaling and growth needs, nClouds recommended OnScale take a serverless approach to its architecture on AWS. “We developed a serverless API for OnScale with (Amazon) API Gateway to deploy compute power for end users,” explains Carlos Rodriguez, DevOps consultant at nClouds. “OnScale isn’t running any servers. They use AWS Lambda and Amazon Elastic Container Registry (ECR) to deploy Docker containers.”

“We use AWS with Docker to deploy solve engines onto as many instances as we want,” says Banks. “We monitor all of the instances, and then take the data back to Amazon Simple Storage Service (S3) where the end user can then push or pull the results. The whole idea is to let AWS take the burden of the computation and the storage and de-escalate it on the user’s side.” It’s a straightforward process for end users to use the OnScale UI, as the complexity involved with running simulations is abstracted away. “Engineers want a seamless experience and results back quickly,” says Campbell. “That experience is what we’re providing them with our solution.”

Diagram 1: The OnScale Cloud Architecture on AWS

Security is the top priority for the OnScale team. OnScale uses Amazon Cognito for user authentication and the AWS Security Token Service (STS) to create temporary credentials for trusted users. OnScale uses AWS data encryption schemas to encrypt data while in transit via SSL with 256-bit AES encryption. Data is also encrypted when stored in private buckets on S3. At no point is sensitive customer data decrypted during transfer or storage. Simulation jobs run in private, closed containers that are inaccessible outside of AWS with hardware-level encryption. Encryption key management is provided by AWS Key Management System (KMS). Encryption keys are created locally when a user first authenticates OnScale on his or her local machine. Keys are never shared with OnScale.

“All of the out-of-the-box components that native AWS services such as Cognito and KMS provide immediately remove a great deal of the worry and work that we’d normally have configuring services,” says Campbell. “We’re then enabled to put more time towards optimizing our solvers.”

By using AWS and providing users a flexible SaaS subscription model, OnScale offers engineers across disciplines nearly infinite HPC compute resources to run complex simulations, for a fraction of the time it’d take and cost to run locally. In one example, which you can view here, OnScale ran 400 simulations in parallel on 400 AWS Cloud containers which took the team just two minutes to complete. Running locally, the team was limited to running the simulations in a series, taking a day to complete all 400 runs. “For engineering firms in highly competitive industries, the benefits of running on AWS are many,” says Campbell. “You can get products modeled, developed, and up-and-running faster which allows you to be more competitive, and you can lower your R&D costs to get your products to market.”

The team is grateful for the AWS migration, DevOps, and security expertise that nClouds brought to the engagement. “It would have been a much longer process for us to develop our architecture on AWS without a partner like nClouds,” says Harvey. “You need cloud engineers to engineer on the cloud and we didn’t have that capability ourselves. My feeling is, left to our own devices, it would’ve taken us at least 3 - 4x the time to get to a workable solution and there would’ve still been an element of greenness on our part, like were we actually using services correctly on the backend, for instance? nClouds has been integral in helping us along the way to solve problems and understand how to optimize our use of different services.” “Our AWS account manager had nothing but glowing things to say about nClouds, and I have to say, he was right in his evaluations of their abilities to help us,” adds Banks.

“With our approach to engineering software and HPC resources, we want to open up brand new opportunities for engineers across markets with a specific focus on IoT sensors, 5G RF connectors, BioMed, and ADAS (driverless cars),” says Campbell. “We’re just getting started.”

OnScale’s SaaS platform on AWS is now generally available. Click here to learn more.

Learn more about DevOps and AWS.