Microsoft and NetApp have shattered performance barriers for cloud-based electronic design automation (EDA) workloads, publishing new SPECstorage Solution 2020 EDA_BLENDED benchmark results on April 24, 2026. The tests reveal that Azure NetApp Files large volume breakthrough mode can scale to 17,280 job sets while maintaining an average overall response time (ORT) of just 0.60 milliseconds. This result cements Azure’s position as a leading platform for chip design, simulation, and verification, where storage latency and scale directly impact time to tape-out.

Electronic design automation is notoriously I/O-intensive. Modern chip designs involve billions of transistors, requiring massive parallel simulations that hammer storage with concurrent reads, writes, and metadata operations. Traditional on-premises parallel file systems like Lustre or IBM Spectrum Scale have long been the go-to, but they demand complex management and overprovisioning to meet performance spikes. Cloud-native alternatives are finally closing the gap, and this benchmark proves it.

The SPECstorage EDA_BLENDED Benchmark Decoded

SPECstorage Solution 2020 EDA_BLENDED is the industry’s hardest storage baseline for EDA workloads. Unlike generic throughput or IOPS tests, it models a complete mix of operations found in real chip design flows: logic simulation, static timing analysis, power analysis, and place-and-route. It stresses the storage system with large-file streaming, small-file random access, and metadata-heavy directory scans—all running concurrently across thousands of job sets.

The metric of merit is not raw bandwidth. It is the number of job sets a storage solution can support while staying below a specified overall response time (ORT) threshold. The benchmark is strict: average latency must be under 40 milliseconds, and 95th percentile latency must also be below a defined ceiling. In practice, the highest-performing systems cluster around an ORT of 1–3 milliseconds. Azure NetApp Files breakthrough mode delivering 0.60 ms at 17,280 job sets is extraordinary—it’s effectively real-time.

Unpacking the Numbers: 17,280 Job Sets at 0.60 ms

To appreciate the scale, consider that one job set roughly corresponds to a single engineer’s interactive simulation session. 17,280 job sets could represent the peak workload of a global semiconductor company running hundreds of parallel regression suites. Each job set continuously issues a stochastic mixture of I/O operations:

  • File creation and deletion: Thousands of temporary checkpoint files per hour.
  • Random reads/writes: Partial netlist and waveform updates.
  • Metadata queries: File stats, directory listings, permission checks.
  • Large sequential reads: Loading cell libraries and design databases.

At 0.60 ms average response time, the storage layer becomes invisible. Engineers experience no lag when launching simulations, viewing logs, or saving results. This low latency also enables deeper parallelization without dead time—critical as EDA tools increasingly adopt cloud-native architectures that expect API-like storage responsiveness.

What Is Azure NetApp Files Large Volume Breakthrough Mode?

Azure NetApp Files is a first-party, bare-metal storage service built on NetApp ONTAP, running natively inside Azure data centers. It combines ONTAP’s enterprise data management—snapshots, cloning, replication, and compression—with the elastic scale of Azure. Historically, a single NetApp volume topped out at 100 TiB. Large volume support, introduced a few years ago, raised that limit to 500 TiB and eventually to multiple petabytes, but performance scaling was not perfectly linear.

Breakthrough mode is a recent innovation that changes the performance profile at petabyte scale. By optimizing the data path, leveraging higher-throughput back-end network interconnects, and fine-tuning ONTAP’s write allocation and caching algorithms, it delivers consistently low latency even as capacity and concurrency soar. Microsoft and NetApp have engineered the service to maintain sub-millisecond latency for random I/O patterns typical of EDA, without requiring customers to manually tune stripe sizes or file system parameters.

Key architectural enablers include:

  • Tiered caching with NVMe-backed read accelerators: Hot metadata and small files are cached locally, reducing back-end traffic.
  • Parallel front-end protocol termination: NFSv3/v4.1 sessions are load-balanced across dozens of service nodes, preventing any single point from becoming a bottleneck.
  • Optimized placement of file system metadata: ONTAP’s inode table is distributed and cached in a way that metadata operations scale linearly with the number of volumes, not with the number of files in a single directory.
  • Direct binding to Azure accelerated networking: RDMA-capable virtual NICs provide microsecond tail latencies between Azure VMs and the storage backend.

Why EDA Demands Extreme Storage Performance

Chip design is a race against time. A single bug caught late in the design cycle can cost millions in mask respins. Regression testing—rerunning all tests after a change—must be completed overnight to keep the next day’s engineering work on track. Storage latency multiplies directly with the number of regressions. Moving from 2 ms to 0.6 ms can reduce a 10-hour regression run by hours, compressing the iteration cycle and accelerating time to market.

Cloud EDA also introduces new dynamics. Unlike on-premises farms where storage is overbuilt for peak demand, cloud workloads scale up and down. If storage latency spikes when 5,000 NFS clients mount the same volume, engineering teams must either live with slower simulations or implement complex job scheduling to avoid contention. Breakthrough mode makes such spikes practically impossible, offering deterministic latency up to extreme loads.

Inside the Test Configuration

While the full result is published on SPEC.org, typical EDA_BLENDED runs on Azure involve a cluster of high-performance VMs—likely HBv4 or HBv5 series with InfiniBand—each simulating multiple job sets. The storage endpoint is an Azure NetApp Files large volume configured with the breakthrough mode feature enabled. The volume is exported via NFSv3, the protocol of choice for most EDA tools.

The published ORT of 0.60 ms is an average across all operations. The fact that it was sustained at 17,280 job sets indicates that the storage system can handle massive parallelism without degrading. For context, previous public results on Azure NetApp Files without breakthrough mode achieved around 10,000 job sets at approximately 1.2 ms. The new result represents a 73% increase in scale and a 50% reduction in latency.

Comparative Landscape: On-Premises vs. Cloud

For decades, EDA storage has been synonymous with dedicated parallel file systems. An on-premises Lustre cluster can achieve similar or slightly better latency at comparable scale, but only after months of procurement, installation, and tuning—and at a cost that includes idle capacity during off-peak times. Cloud storage delivers elasticity: you pay for the capacity you use, and performance scales with provisioned bandwidth.

Azure NetApp Files’ breakthrough mode further narrows the gap. It provides an enterprise-grade, fully managed NFS service that can back thousands of concurrent mount points without a dedicated administrator. Data protection features like scheduled snapshots, cross-region replication, and instant zero-footprint clones add critical value for EDA workloads, where a golden design database must be preserved and branched efficiently.

Competitors like AWS FSx for NetApp ONTAP or Google Cloud NetApp Volumes offer similar NetApp-powered capabilities, but Azure’s tight integration with the ONTAP platform and custom hardware acceleration gives it a performance edge in this benchmark. The SPEC result is a testament to the co-engineering between Microsoft and NetApp.

Customer Impact: Faster Time to Tape-Out

Large semiconductor companies, fabless design houses, and EDA software vendors stand to benefit immediately. For example, a company running Synopsys or Cadence tool flows can now consolidate multiple NAS gateways into a single Azure NetApp Files volume, reducing administrative overhead. They can burst to 17,000 concurrent simulation jobs without pre-warming caches or worrying about noisy neighbors.

Key operational benefits:

  • Simplified architecture: No need for a parallel file system overlay on top of cloud VMs. Native NFS serves the entire workload.
  • Instantaneous space-efficient cloning: Using NetApp’s FlexClone, engineers can spin up per-branch copies of multi-terabyte design databases in seconds, accelerating parallel development.
  • Global collaboration: With cross-region replication, geographically dispersed teams work on the same dataset with minimal latency overhead.
  • Cost control: The service’s reserved capacity options and zero-commit Flex ability let organizations align storage spending with project lifecycles.

The Path Forward for Cloud-Native EDA

This benchmark is more than a vanity metric. It signals that public cloud storage has matured to handle the planet’s most demanding engineering workflows. The shift from on-premises to cloud for EDA is accelerating, driven by the need for scalable compute and collaborative design. Storage must not be the bottleneck.

Microsoft and NetApp have indicated that further optimizations are in the pipeline, including support for NFS over RDMA with the next-generation ONTAP release, which could push latency below 0.4 ms. Moreover, Azure’s upcoming Cobalt 2 ARM-based VMs and enhanced networking will likely reduce client-side latency as well.

For EDA teams, the message is clear: the cloud can now match or exceed the performance of bespoke on-premises storage, without the management burden. The SPEC EDA_BLENDED result for Azure NetApp Files breakthrough mode is a landmark that will influence procurement decisions for years to come.

The full result is available on the SPEC website.