How Silicon Labs accelerates IoT innovation at scale

Silicon Labs is a semiconductor company specializing in secure, low-power chips and software for Internet of Things (IoT) devices. Headquartered in Austin, Texas, they operate globally with offices across North America, Europe, and Asia. They design microcontrollers, wireless systems-on-chip, and development tools that help manufacturers build connected products efficiently and reliably. Known for their focus on energy efficiency, security, and robust wireless performance, Silicon Labs plays a significant role in enabling smart home, industrial automation, and commercial IoT solutions. The company positions itself as an innovation-driven, engineering-focused leader in the connected device ecosystem.

Silicon Labs is a semiconductor company headquartered in Austin, Texas, with locations worldwide. They design chips that power billions of Internet of Things (IoT) devices, from wirelessly controlled light bulbs to door locks and sensors, connecting everyday appliances to the cloud for a smarter and more energy-efficient world.

The semiconductor design process relies heavily on electronic design automation (EDA) tools that are extremely data intensive. "EDA workloads are very heavily metadata-oriented. They tend to be in the 70% to 85% metadata from an I/O distribution standpoint," explains Mike Webb, Staff Software Engineer at Silicon Labs. Each project generates billions of small files and directories, creating massive performance bottlenecks for traditional storage systems.

"We use extensive simulation in our design process to verify that our chips are rock solid before they go further down the chip design pipeline," Webb explains. When systems fail, the impact is immediate and costly. "When things don't work, if there's an outage or something, for our designers, you're losing millions of dollars. It delays everything."

As chip design complexity grew alongside emerging AI initiatives and expanding simulation requirements, Silicon Labs needed storage solutions that could deliver high performance, reliability, and simplified management, especially at remote sites without dedicated IT staff.

Silicon Labs' existing storage infrastructure was stretching to meet the demands of modern chip design. Legacy systems required complex configurations and careful provisioning across back-end disks to handle the metadata-heavy workloads.

The challenge intensified as the company expanded its use of AI for design verification and simulation. Webb, Staff Software Engineer on the Cloud Software Engineering Team, was building AI-powered knowledge search tools to help designers access information faster. "Gen AI is very exciting," Webb explains. "When a new technology like this comes out, all these different valuable opportunities emerge."

Webb was initially drawn to cloud providers for AI development, but quickly encountered limitations. "When using our previous cloud provider for this storage for these AI workloads, we had bandwidth limitations because you're talking about a lot of training data," he notes. The bandwidth constraints slowed iteration cycles and hindered application development.

Beyond performance issues, cloud storage raised security and cost concerns for sensitive design data. Silicon Labs needed an on-premises solution that could support their needs while maintaining the performance and security required for proprietary chip designs, and they looked to Everpure FlashBlade® and Portworx®.

Silicon Labs, using the Everpure platform, plans to build an Enterprise Data Cloud to unify all of their data, structured and unstructured, on premises and in the cloud. They are using FlashBlade to create a unified pool of storage that serves as the backbone for their EDA workloads and simulation, design, and AI/ML workloads. The company also implemented FlashArray™ for block-based production workloads including SQL databases and virtual machines.  Additionally, Silicon Labs uses Portworx to enable modern cloud-native orchestration and automate key data protection tasks, such as cloud-based snapshots and scheduling, supporting their high-performance engineering and AI workloads.

"We took, I think it was three-and-a-half cabinets that were 52U in height that were entirely full of another storage vendor's storage and basically cut it in half," explains Webb. The new system not only used substantially less space, but also delivered superior performance while consuming significantly less power after removing approximately 1,300 spinning disks.

"Some simulations are hundreds of terabytes, and they can be run directly from FlashBlade," notes Webb. The high-speed, local storage enabled his team to build an in-house AI knowledge search engine using the FlashBlade native S3 endpoint. "Having this on-prem flash storage has enabled me to really do more experiments quicker without having to worry about cloud costs or data privacy issues,” says Webb. “It allowed us to iterate faster on our design for our AI workloads by just having so much quick compute right next to where the application's developed."

The S3 compatibility proved especially valuable for developers. "Because at the end of the day, deadlines are tight, and you don't want to have to change your whole workflow for your storage vendor," says Webb. The system allows developers to build sophisticated AI applications without infrastructure constraints, helping designers quickly access disparate information sources including design guidelines, references, and design reviews, accelerating the entire development process.

The results exceeded expectations across multiple areas. Silicon Labs achieved performance gains of up to 40 times faster for some simulation workloads compared to legacy infrastructure while supporting massive simulations measuring hundreds of terabytes and tens of thousands of design iterations.

The Everpure Evergreen//Forever™ subscription enables nondisruptive upgrades and continuous access to the latest technology without costly refresh cycles. By consolidating EDA, business, and manufacturing workloads on the same FlashBlade system, Silicon Labs reduced management complexity while supporting future scaling needs. Complementing this foundation, Portworx streamlines data protection and storage management across the environment, delivering automated workflows and high IOPS scratch space for simulations that accelerate product innovation and reduce time to market for IoT devices.

"Having reliable infrastructure on premises with something like FlashBlade allows our software developers to work on the problems that are truly valuable to the organization and not be distracted by an unplanned outage or configuration change," says Webb.

The company reduced rack space for core workloads by 85% and halved the cost per terabyte while increasing speed and reliability. Power consumption also dropped substantially, supporting Silicon Labs' sustainability goals.

"We don't have a lot of issues," notes Webb. Since deployment, Silicon Labs has reported zero production outages. For Webb, the impact is clear: "It helps me do my job and to do what really matters, which is building these devices that make a difference."