Azure NetApp Files for EDA workloads: From revolution to breakthrough at scale

Last year, we outlined how Azure NetApp Files helped reshape silicon design by delivering the low-latency, high-throughput storage required for Electronic Design Automation (EDA) workloads at cloud scale. Since then, we have continued to extend performance and scalability. Today, we are advancing that progress with another significant step forward.

Modern semiconductor design is defined by scale. Thousands of concurrent EDA jobs spanning simulation, synthesis, and verification run continuously against shared datasets, where even small variations in storage latency can ripple across entire design cycles. For many teams, this has historically limited how far EDA workflows could scale in the cloud.

That constraint is now changing.

Azure NetApp Files (ANF) is redefining what is possible for EDA in the cloud by delivering predictable, high-performance shared storage at massive concurrency. With new independent benchmark results and growing adoption by leading semiconductor companies, Azure is establishing itself as a viable—and in many cases superior—platform for modern EDA environments.

See how Azure NetApp Files delivers scalable, high-performance storage for EDA

Why EDA storage has been difficult to scale in the cloud

EDA workloads combine three characteristics that have traditionally challenged cloud storage architectures:

Extremely high concurrency, with thousands of jobs accessing shared file systems simultaneously.

Strict latency sensitivity, where even minor delays reduce compute efficiency and extend runtimes.

Intensive shared data access patterns, creating contention under load.

While cloud compute scales easily, shared storage has often introduced variability that limits overall system efficiency. As concurrency increases, storage becomes the bottleneck, impacting regression cycles, increasing tool license costs, and slowing time to tape-out.

For EDA teams evaluating cloud transformation, the central question has remained consistent: can storage scale with compute while maintaining predictable performance?

A modern approach: Azure NetApp Files for EDA at scale

Azure NetApp Files is designed specifically to address this challenge. Its architecture aligns directly to the requirements of highly parallel, shared workloads like EDA.

At its core, ANF enables independent scaling of compute and storage, so EDA clusters can grow without storage becoming the constraint, and additional compute nodes do not introduce hotspots or contention at the storage layer. It natively supports concurrent metadata operations at scale, handling the millions of small file interactions typical of EDA workflows without degradation. And its service-level performance model ensures that throughput and IOPS scale predictably with capacity, eliminating the need for complex tuning.

More recently, innovations such as large volumes and large volumes breakthrough mode have expanded the concurrency envelope even further. These capabilities allow thousands of parallel jobs to share a single storage environment while maintaining consistent latency under sustained load.

This delivers what cloud-based EDA systems have long struggled to provide: consistent, repeatable performance, not only at low utilization, but also under full production load.

Independent validation: SPECstorage® Solution 2020 benchmark results

To validate these capabilities in a real-world context, Azure NetApp Files was measured using the industry-standard SPECstorage® Solution 2020 EDA_BLENDED benchmark. This benchmark simulates realistic EDA workflows by combining metadata-intensive frontend operations with throughput-heavy backend processing, all under strict latency requirements.

The Azure NetApp Files large volume breakthrough mode scale configuration reached 17,280 SPECstorage® Solution 2020 EDA_BLENDED JOBS with an overall response time of 0.60 milliseconds (ms).

These results demonstrate several important characteristics:

The ability to sustain very high levels of concurrent EDA workloads.

Consistently low response times under load.

Linear scaling behavior as concurrency increases.

No requirement for overprovisioning.

Historically, top benchmark outcomes in this category have been associated with tightly integrated on-premises systems. This validation underscores a broader shift in the industry: when architected correctly, cloud-based EDA infrastructure can not only match on-premises approaches, but in some scenarios surpass them in both scale and operational efficiency.

Proven in production: EDA workloads already running on ANF

This performance is not limited to benchmarks. Organizations such as AMD and ASML are already using Azure NetApp Files to run EDA and high-performance design workloads in production environments.

These companies operate at the leading edge of semiconductor innovation, where infrastructure must support both extreme scale and precise predictability. Their adoption of ANF reflects a broader industry trend: moving EDA workloads to the cloud is no longer experimental, it is becoming a strategic advantage.

These customers, along with others, consistently report the same operational benefits:

The ability to increase regression concurrency without performance degradation.

Improved utilization of compute resources and reduced EDA tool license fees.

Greater predictability in design cycles, enabling more confident scheduling of key milestones.

In this context, storage is no longer the limiting factor—it becomes an enabler of scale.

How Azure helps EDA teams scale with confidence

Organizations have flexibility in how they deploy EDA environments with Azure NetApp Files, depending on workload characteristics and operational priorities.

Some teams choose a centralized model built around a single large volume to maximize throughput and tightly control latency. Others adopt a multi-volume approach to distribute workloads and scale concurrency across different job types. Many enterprises extend existing on-premises environments into Azure, using cloud capacity to absorb peak demand without permanent infrastructure expansion.

Across all of these patterns, one principle remains consistent: storage performance must scale predictably alongside compute. Azure NetApp Files provides that foundation.

Azure NetApp Files delivers the consistent, high‑throughput NFS performance that modern EDA workloads demand, shrinking runtimes, accelerating tape‑out schedules, and giving chip designers the confidence that storage will never be the bottleneck.
Srikanth Gubbala, Head of Global HPC Infrastructure, Applied Materials

Bringing it all together

The evolution of cloud storage for EDA marks an important inflection point for the semiconductor industry. What was once considered a tradeoff—scale versus predictability—is no longer a constraint.

With Azure NetApp Files, organizations can confidently run highly concurrent EDA workloads in the cloud, supported by architecture designed for their specific demands and validated by independent benchmarking.

For teams exploring how to modernize their EDA infrastructure, the path forward is increasingly clear. Cloud-based storage can now meet the requirements of even the most demanding design environments, while offering the flexibility to scale as workloads continue to grow.

For a deeper technical exploration of the benchmark configuration and design considerations, see the companion Azure Tech Community technical blog: “From scale to breakthrough: Azure NetApp Files sets a new cloud benchmark for EDA.”

For further information, explore the Azure NetApp Files documentation or email askanf@microsoft.com.

Scale high-performance EDA workloads with Azure NetApp Files
Discover how Azure NetApp Files delivers predictable, high-performance storage for EDA workloads, enabling massive concurrency, low latency, and consistent scaling in production.

Get started with ANF

The post Azure NetApp Files for EDA workloads: From revolution to breakthrough at scale appeared first on Microsoft Azure Blog.
Quelle: Azure

Powering multi-cluster workloads with seamless cross‑cluster networking for Azure Kubernetes Fleet Manager

In this article

The challenge of multi-cluster networkingOur vision: Multi-cluster management with seamless networkingStrategic resilience with cross-cluster networkingGetting started with cross-cluster networkingDocumentation and resources

As organizations modernize their application portfolios, we are witnessing a fundamental shift in how cloud-native infrastructure is architected. No longer is the question “How do we scale a cluster?” but rather “How do we scale across clusters, regions, and clouds while maintaining operational simplicity?”

Today, we are thrilled to announce the public preview of cross-cluster networking for Azure Kubernetes Fleet Manager. This capability represents the next evolution in multi-cluster management by introducing transparent E-W multi-cluster networking powered by Advanced Container Networking Services.

Learn more about cross-cluster networking for Azure Kubernetes Fleet Manager

The challenge of multi-cluster networking

Whether driven by regulatory requirements, regional disaster recovery, or the need to isolate blast domains, organizations of every size often run multiple Azure Kubernetes Service (AKS) clusters. However, managing these clusters has historically introduced a “networking tax.” Traditional approaches rely on complex VPNs, gateways, and manual service discovery, adding latency and operational complexity.

Even when operating just a few clusters and especially when operating large scale fleets of clusters, teams need consistent, reliable cross‑cluster connectivity to support scenarios like failover, shared services architectures, and seamlessly shifting workloads across regions for capacity or latency. At the same time, platform teams want to abstract infrastructure details from developers, enabling seamless cluster-level changes without disrupting applications.

Our vision: Multi-cluster management with seamless networking

In response to similar challenges, we built Azure Kubernetes Fleet Manager. Fleet Manager is designed to simplify multi-cluster Kubernetes for everyone. While Fleet Manager has already simplified workload propagation (deploying to many clusters) and update orchestration (safe, staged upgrades), the network remained a challenge.

With the introduction of Cilium-based cross-cluster networking in Azure Kubernetes Fleet Manager, we are delivering a managed, high-performance network that can span your entire fleet.

This capability extends the Kubernetes networking model across clusters, enabling services and workloads to communicate across cluster boundaries as if they were local, while preserving cluster-level isolation and governance.

Built on an open-source foundation, this capability uses Cilium for dataplane and Kubefleet for fleet-level orchestration, both active Cloud Native Computing Foundation (CNCF) projects. This ensures transparency, portability, and alignment with the broader Kubernetes ecosystem, while benefiting from continuous innovation from the open-source community.

The following diagram shows how clusters in a fleet are connected through a unified, managed network, enabling seamless communication, service discovery, and policy enforcement.

Key capabilities include:

Seamless east-west connectivity: Using eBPF-based routing with power of Azure CNI powered by Cilium and Advanced Container Networking Services, pods can communicate across clusters with native performance, no proxies or gateways required.

Global service discovery: With a simple annotation (service.cilium.io/global=true), a standard Kubernetes Service becomes “global.” Cross-cluster networking automatically discovers endpoints across joined member clusters, providing transparent load balancing and failover.

Multi‑cluster observability: Gain a unified view of network health across clusters with aggregated metrics, logs, and flow visibility. Advanced Container Networking Services integrates Cilium telemetry to provide consistent insights, faster troubleshooting, and end‑to‑end visibility across the fleet.

Unified security and governance: Security policies are no longer confined by cluster boundaries. Through Advanced Container Networking Services, you can now enforce enterprise-grade network policies and gain deep observability across your entire global footprint, ensuring identity-based security follows your workloads wherever they run.

Zero-touch management: Fleet Manager handles the complex lifecycle, managing certificates, and network configurations, so you don’t have to.

These capabilities are using eBPF to enable efficient routing, policy enforcement, and observability for high-performance networking

Strategic resilience with cross-cluster networking

In a digital-first economy, resilience is a competitive advantage. Cross-cluster networking enables customers to build architectures that are inherently resilient to single-cluster or single-region failures.

Whether you are running “Shared Services” clusters to support hundreds of tenants or building “Global Services” that route traffic to the healthiest available endpoints, cross-cluster networking for Azure Kubernetes Fleet Manager ensures your infrastructure is as agile as your business needs.

We are committed to providing the most robust, secure, and performant platform for multi-cluster environments. Cross-cluster networking is a big step towards a future where the physical boundaries of a cluster no longer limit the innovation within it.

Getting started with cross-cluster networking

Cross-cluster networking for Azure Kubernetes Fleet Manager is designed to minimize operational complexity:

Prerequisites for your clusters:

Azure CNI powered by Cilium as the networking dataplane.

Advanced Container Networking Services enabled.

Set up cross-cluster networking:

Join clusters to a Fleet.

Associate the members with a cross-cluster network profile.

Deploy services with global annotations to enable cross-cluster communication.

Once configured, Fleet Manager automatically deploys and manages the required components, enabling direct pod-to-pod communication across clusters without additional gateways or overlays.

This managed approach removes the burden of setting up and maintaining Cilium multi-cluster components manually, allowing teams to focus on application delivery instead of infrastructure management.

See it in action: Watch the Cross-Cluster Networking for Azure Kubernetes Fleet Manager Video Guide to learn more and see a quick demo.

Documentation and resources

Cross-cluster networking for Azure Kubernetes Fleet Manager

How to enable cross-cluster networking for Azure Kubernetes Fleet Manager

Azure CNI powered by Cilium

Advanced Container Networking Services

If you have feedback or would like to learn more, reach out to your Microsoft account team or share feedback through the Azure Kubernetes Service community channels, we would love to hear from you!

Get started with cross-cluster networking

The post Powering multi-cluster workloads with seamless cross‑cluster networking for Azure Kubernetes Fleet Manager appeared first on Microsoft Azure Blog.
Quelle: Azure