What is NetApp Azure Files?
Native to Azure, NetApp Azure Files provides enterprise-class, high-performance file storage. It offers Volumes as a Service, which you may generate in a capacity pool with a NetApp account and distribute to clients via SMB and NFS. Additionally, you may control data protection and choose service and performance levels. The same protocols and tools you know and use on-premises may be used to construct and administer scalable, high-performance, and highly available file shares.
NetApp Azure Files primary characteristics are:
- Scale, cost optimisation, and performance.
- Availability and ease of use.
- Data security and management.
Azure NetApp Files may be utilised for the following use cases and supports SMB, NFS, and dual protocols volumes:
- Sharing of files.
- Directories at home.
- Databases.
- Computing with high performance.
HPC workloads strain cloud infrastructure, requiring scalable and dependable resources to manage complex computational tasks. These workloads require a lot of parallel processing capacity, which is generally provided by CPU or GPU virtual machine clusters. HPC applications demand fast access speeds and big data storage, which cloud file systems cannot provide. To satisfy the requirements for high throughput input/output (I/O) and low latency, specialised storage solutions are needed.
Low latency, excellent speed, and enterprise-grade data management at scale are all features of Microsoft Azure NetApp Files. Because of its special features, NetApp Azure Files may be used for a variety of high-performance computing tasks, including risk modelling, seismic processing, reservoir simulations, and electronic design automation (EDA).
Infrastructure requirements of EDA workloads
In order to handle complicated tasks in simulation, physical design, and verification, EDA workloads demand a high level of computing power and data processing. Multiple simulations are used at each design step to increase accuracy, dependability, and the early detection of design flaws, which lowers the cost of debugging and redesigning. Additional simulations may be used by silicon development engineers to evaluate various design scenarios and optimise the Power, Performance, and Area (PPA) of the device.
Frontend and backend workloads are the two main categories of EDA workloads, and each has specific needs for the underlying computation and storage infrastructure. Frontend workloads comprise thousands of short-duration parallel operations with an I/O pattern that is characterised by frequent random reads and writes across millions of small files, with an emphasis on logic design and functional elements of chip design. Backend workloads, which comprise hundreds of processes using the sequential read/write of fewer, bigger files, are centred on converting logic design to physical design for production.
It is not easy to choose a storage solution to accommodate this particular combination of frontend and backend workload patterns. The SPEC SFS benchmark was created by the SPEC collaboration to aid in comparing the many storage options available in the market. The EDA_BLENDED benchmark offers the distinctive patterns of front-end and back-end workloads for EDA workloads. The following table provides a description of the composition of I/O operations.
| EDA workload stage | I/O operation types |
| Frontend | Stat (39%), Access (15%), Read File (7%), Random Read (8%), Write File (10%), Random Write (15%), Other Ops (6%) |
| Backend | Read (50%), Write (50%) |
Regular volumes, which are perfect for applications like databases and general-purpose file systems, are supported by NetApp Azure Files. Multiple regular volumes are necessary for EDA workloads, which operate with massive amounts of data and demand very high throughput. For EDA applications, the advent of big volumes to accommodate larger data amounts is beneficial since it streamlines data management and provides better performance than several normal volumes.
The results of the SPEC SFS EDA_BLENDED benchmark performance tests are shown below, showing that NetApp Azure Files can provide ~10 GiB/s throughput with less than 2 ms latency when employing big volumes.
Electronic Design Automation at Microsoft
Microsoft is dedicated to integrating Artificial Intelligence into all tasks and user interfaces for both current and future devices. The design and production of silicon is the first step. By using Azure for it own chip design requirements, Microsoft is pushing the bounds of Moore’s Law and surpassing scientific barriers at a rate never seen before in the history of EDA processes.
Some of Microsoft’s first entirely tailored cloud silicon chips were developed in large part because to the application of the best practices approach to optimise Azure for chip design with suppliers, partners, and customers:
- The Azure Maia 100 AI Accelerator is designed for generative AI and AI activities.
- An Arm-based processor designed to do general-purpose computing tasks on Microsoft Azure is the Azure Cobalt 100 CPU.
- Microsoft’s latest proprietary security chip, the Azure Integrated Hardware Security Module, is intended to strengthen key management.
- The Azure Boost DPU is the company’s first internal data processing unit, built for low-power, high-efficiency data-centric applications.
In addition to providing best-in-class computational capabilities for HPC workloads, the chips created by the Azure cloud hardware team are installed in Azure servers, further accelerating the rate of innovation, dependability, and operational efficiency utilised to construct Azure’s production systems. The Azure cloud hardware team benefits from using Azure for EDA in the following ways:
- Quick access to state-of-the-art, scalable, on-demand processors.
- Each EDA tool is dynamically paired with a particular CPU architecture.
- Using Microsoft’s advancements in AI-powered semiconductor workflow technology.
How NetApp Azure Files accelerates semiconductor development innovation
Superior performance
NetApp Azure Files can achieve 826,000 IOPS at the performance edge (~7 milliseconds of latency) and up to 652,260 IOPS with less than 2 milliseconds of latency.
High scalability
The amount of data produced might increase dramatically as EDA initiatives progress. Large-capacity, high-performance single namespaces with volumes up to 2PiB are offered by NetApp Azure Files, which also scales easily to accommodate compute clusters with up to 50,000 cores.
Operational simplicity
Simple and easy to use, Azure NetApp Files may be accessed using the Azure Portal or an automated API.
Cost efficiency
Cool access is provided by Azure NetApp Files, which allows cool data blocks to be transparently moved to the managed Azure storage tier for lower costs. Upon access, the hot tier is instantly restored. The high prices of enterprise-grade storage solutions are further decreased by NetApp Azure Files reserved capacity, which offers considerable cost reductions over pay-as-you-go pricing.
Security and reliability
With key management and encryption for both data in transit and data at rest, Azure NetApp Files offers enterprise-grade data management, control-plane, and data-plane security capabilities that guarantee the availability and protection of vital EDA data.
The Azure cloud hardware team built a production EDA cluster on Azure, where NetApp Azure Files provides clients with more than 50,000 cores per cluster, as seen in the graphic below.
