Executive Interview: Ian Campbell founder and CEO of OnScale
Do your simulation in the cloud
OnScale is a relatively new company; can you tell us how the company got started?
OnScale is the combination of a seasoned multiphysics solver team with Silicon Valley tech entrepreneurs. The solvers have their roots decades ago and were developed and maintained by Weidlinger Associates/Thornton Tomasetti, one of the largest engineering consulting firms in the world. In 2017, Thornton Tomasetti spun out the solver team and recruited a group of Silicon Valley entrepreneurs to put the solvers onto the cloud and create OnScale, which is exactly what the team did. In 2018, OnScale secured additional funding from Intel Capital and Google’s Gradient Ventures, and now, both Intel and Google are OnScale customers and partners.
What technology is the basis of the company’s software solvers and fast computation capabilities?
The core solver technology is called FLEX. The FLEX code base was developed at the same time as other solvers like Abaqus and LS-Dyna, although FLEX was used exclusively at Weidlinger/Thornton Tomasetti for projects for the DOD, DARPA, other U.S. and global defence agencies, and hundreds of Fortune 100 companies. FLEX was consciously architected to be extremely RAM and compute cycle efficient. When new multiphysics components were needed, they were carefully grafted onto the FLEX core in a way that allowed the code to be easily parallelized, creating a very tightly coupled multiphysics solver core. Capabilities like MPI (Message Passing Interface) were added to aid in massive scalability across DOD supercomputers with thousands of cores, allowing users to solve very large real-world problems. This architecture worked extremely well in the age of mainframe computing, and is a natural fit for modern cloud computing which is extremely parallel, giving OnScale the ability to scale-out across thousands or even hundreds of thousands of cores on the cloud.
What are the advantages of cloud computing for modeling of complicated circuits?
The general advantages of cloud computing are two-fold:
- Ability to run extremely large simulations (on the order of billions of degrees of freedom) thus eliminating the number of approximations that have to be made when physical phenomena is translated into a simulation.
- Ability to run extremely large sets of simulations in parallel for parametric sweeps, DOEs, optimization studies, and Monte Carlo studies.
Essentially, the cloud is the biggest supercomputer ever built. OnScale allows any engineer to access this supercomputer. Specifically, for circuit analysis, this allows engineers to study extremely complicated circuits coupled with physical components like SAW (Surface Acoustic Wave) and BAW (Bulk Acoustic Wave) filters. OnScale customers routinely run full die simulations of multiple resonator circuits (duplexes, quadplexers, etc) coupled to other RF front-end components like antennas and PAs. More on this a little later.
How are the IP and design models of the customer protected in a cloud-based system – are there specific security measures in place?
Absolutely! OnScale is built entirely using Amazon Web Services (AWS) and Google Cloud Platform (GCP) components. Things like identity management and encryption are handled entirely by these world-class services and are much more advanced than many local data centers, which are prone to phishing hacks and physical hacks, to cite a few. When you log into OnScale, you’re actually logging into AWS via AWS Cognito. If you lose your password or encryption key, you have to get it from AWS via the OnScale interface. No one at OnScale can see any of your encrypted data on AWS or GCP. No one else without your encryption key can see any of your data. To hack your data on AWS, a hacker would (A) have to know what data center you’re using, (B) know which server(s) you’re using inside that data center, (C) ninja-style rappel into the data center after evading AWS physical security, which is impossible, and (D) pause your simulation and do a RAM dump, since everything is encrypted on the file system. In other words, it’s virtually impossible to hack such a cloud architecture.
Is there a cost or time to market advantages for this type of simulation service versus traditional platforms?
Absolutely. That’s why we started OnScale! Many of us are engineers, and some of us are hardware entrepreneurs who suffered the pains of using legacy simulation tools and were reliant on physical prototyping of things like MEMS sensors. Physical prototyping is enormously expensive in markets like MEMS and RF, some would say it’s unsustainable. OnScale was formed to shift much of the time, cost, and risk of physical prototyping to virtual prototyping and virtual qualification of designs. Not just the die design, but packaging, assembly, integration into systems: many things design engineers don’t think about since they couldn’t digitally prototype them before OnScale. We have many Fortune 100 companies on our customer list who collectively save tens of millions of dollars every year using OnScale, versus cutting wafers and building physical prototypes. Because OnScale is multi-tenant with a consumption based pricing, simulating in the cloud with OnScaleallows new areas, like process engineering, to benefit from virtual prototyping and creation of Digital Twins.
What solutions and models does OnScale have in the area of SAW, BAW, and FBAR filters?
We provide a number of SAW/BAW and FBAR specific Product Development Kits with Simulation Guides https://onscale.com/modeling-guides/ . OnScale has also developing integrations with MATLAB and other EDA tools to rapidly enhance engineering workflows for such devices, which will be critical for developing and bringing to market new 5G devices.
Do you support mechanical and thermal simulations as well and how are they integrated with the electrical simulations?
Yes! Mechanical and thermal simulations are 100% coupled with piezoelectric and wave propagation simulations critical for SAW and BAW device design. This allows engineers to run “virtual qualification” studies simulating the effects of environmental stress tests, primarily thermal stress tests, among many other possible studies.
Do you plan to support other high frequency device design?
Yes! OnScale is proud to announce the availability of the Electromagnetic (EM) simulation capability for high-frequency designs like 5G patch antennas.
What kind of plans are available to users and how do they differ from traditional platform purchases?
Traditionally, simulation software was sold on a one-engineer/one-license/one-computer model. That model just doesn’t make sense in the age of cloud computing, where engineers can literally access thousands of high-performance machines at once. OnScale is sold on a pure “Core-Hour” on-demand, scalable model. If you use 1 core for one hour, that’s a core-hour. If you use 1,000 cores for one hour, that’s a thousand core-hours. OnScale helps engineers efficiently estimate the amount of core-hours they will need for individual simulation studies and on an annualized basis, before they run their simulation.
Our core-hour subscription plans get cheaper the more you simulate. If you sign-up for OnScale’s free service, you get 10 core-hours for free each month: that’s plenty of horsepower to solve interesting 2D and small 3D SAW/BAW simulation problems. If you need more, you can purchase core-hours on-demand at $10 per core-hour. If you’re an enterprise company looking to switch your entire team from legacy CAE simulation tools, then OnScale offers very cost-effective Enterprise subscription plans.
What are your future plans for extending the platform and service?
Our plan is simple: add more features and become the simulation platform of choice for every engineer. Since there is no licenses, add-ons, or modules to buy, every time we release a new feature or a new solver, they’ll become instantly available to every user. OnScale is the Future of Engineering, and our team, our investors, and our customers want us to dominate in every engineering application, from RF and MEMS to IoT, ADAS, Biomed, and sSemiconductor.