Azure NetApp Files: High-Performance Storage for EDA Workloads

The Challenge of EDA Workloads in the Cloud

Electronic Design Automation (EDA) workloads present significant architectural challenges for cloud storage due to their demanding characteristics. They require extremely high concurrency with thousands of jobs accessing shared file systems simultaneously, are strictly latency-sensitive where minor delays reduce compute efficiency, and involve intensive shared data access patterns that create contention under load. While cloud compute scales easily, shared storage has historically been a bottleneck, impacting regression cycles and increasing costs.

Azure NetApp Files Architectural Solution

Azure NetApp Files (ANF) is designed to specifically address these challenges. Its core architecture enables independent scaling of compute and storage, preventing storage from becoming a constraint as EDA clusters grow. This design ensures that adding compute nodes does not introduce hotspots or contention at the storage layer.

• Concurrent Metadata Operations: Natively supports millions of small file interactions typical of EDA workflows without performance degradation.
• Predictable Performance Scaling: Throughput and IOPS scale predictably with capacity, eliminating complex manual tuning.
• Large Volumes & Breakthrough Mode: Recent innovations expand concurrency, allowing thousands of parallel jobs to share a single storage environment while maintaining consistent latency under sustained heavy load.

Validated Performance and Real-World Adoption

ANF's capabilities have been independently validated using the industry-standard SPECstorage® Solution 2020 EDA_BLENDED benchmark, achieving 17,280 concurrent jobs with a 0.60 ms response time. This demonstrates its ability to sustain very high concurrent workloads with consistently low response times and linear scaling. Leading semiconductor companies like AMD and ASML are already leveraging ANF in production, confirming its effectiveness in critical design environments, enabling increased regression concurrency, improved compute utilization, and greater predictability in design cycles.