Arm Scales Performance for Chip Design Using Amazon FSx for NetApp ONTAP

Founded in 1990, Arm is a semiconductor and software design company that designs energy-efficient CPU and GPU processors and system-on-a-chip infrastructure and software. EDA is a crucial component of Arm’s chip-design process; this application encompasses over 50 specialized software tools for the entire lifecycle of semiconductor design.

EDA workloads involve many files and require intensive metadata operations for efficient processing. Because of this complexity, Arm required scalable compute and storage to manage and accelerate the design process.

“The major challenge with high-performance computing workloads, including EDA, is that they are spiky,” says David Miller, senior director and fellow in IT architecture at Arm. “We have times where compute is unused and times where we do not have enough compute to satisfy demand.”

Arm’s original setup included an on-premises data center with NetApp ONTAP storage systems. However, the company needed a dynamic, scalable solution that could handle the variability of its compute needs while providing the agility to scale up or down as required.

Over the years, Arm has developed a strong history on AWS, and, after engaging the AWS team, the company identified Amazon FSx for NetApp ONTAP as an ideal solution. With Arm’s on-premises environment already using NetApp ONTAP, this service aligned with its needs.

Using Amazon FSx for NetApp ONTAP, Arm now runs EDA using a hybrid cloud setup. During periods of peak demand, it bursts workloads to the cloud. This fully managed cloud service replicates the data management capabilities of Arm’s on-premises environment while benefiting from the scalability and flexibility of the cloud. Amazon FSx for NetApp ONTAP offers an ideal pathway to migrate, back up, or burst applications to AWS without necessitating changes to application code or data management practices.

In the hybrid setup, Amazon FSx for NetApp ONTAP is configured as an in-cloud cache for Arm’s on-premises NetApp file system using NetApp FlexCache. With NetApp FlexCache volumes, Arm can burst to the cloud using  Amazon Elastic Compute Cloud (Amazon EC2)—which offers broad and deep compute infrastructure—to provide low-latency access to tens of petabytes of data and tens of billions of files. Arm adopts Amazon FSx for NetApp ONTAP as a cloud-based cache for its on-premises data so that cloud computing workloads can efficiently access data stored on AWS. This setup enhances performance by removing the need for each workload to access on-premises data directly.

To maintain a consistent working experience, Arm also established a cloud foundation platform on AWS, where each workload is separated into its own dedicated environment. This environment supports lightweight directory access protocol, which is used for directory services authentication, and NFSv3, which facilitates file sharing. This setup mirrors the on-premises experience, minimizing disruptions to users’ EDA workloads and effectively replicating Arm’s existing operations in the cloud.

The transition to AWS provided multiple benefits to Arm. Since implementing a modern, cloud-based architecture with Amazon FSx for NetApp ONTAP, Arm has processed 10 million jobs. The company has also enhanced its EDA workflows by using AWS for burst-compute capacity. With this hybrid approach, Arm can complement its on-premises compute with highly scalable cloud resources. These capabilities—combined with the adoption of services such as AWS Batch, a fully managed batch computing service—have accelerated the chip-design process and, thus, its speed to market.

The fully managed nature of the solution significantly reduces operational overhead for Arm’s development and engineering teams. Before Amazon FSx for NetApp ONTAP introduced scale-out features, Arm relied on a single pair of high-availability file servers and horizontal scaling to manage its workload demands. This approach was sufficient for smaller projects, which required one cluster, but scaling larger and more compute-intensive ONTAP workloads was complex and time consuming.

Using scale-out Amazon FSx for NetApp ONTAP, Arm has surpassed the performance of its previous deployments. This feature removes the limits on the number of compute nodes that can be used, so Arm no longer needs to search for compute capacity to support its storage needs. “Some of our workload processing times improved by more than 50 percent,” says Miller. “Amazon FSx for NetApp ONTAP is a game changer for us, and we intend to keep pushing its boundaries.”