The accelerated innovations in semiconductor design have raised customers' expectations of getting smaller, faster, high-quality SoCs at lower costs. However, as a result, the SoC designs are getting complex, resulting in intricate design simulations and explosive data generation. To keep up, the industry is migrating to public/private clouds. Nonetheless, this migration may pose cost/time-to-market issues, leading to further complications. Managing complex data and workflows while maintaining efficiency has become a significant challenge for EDA designers.

Simply Data Synchronization in Hybrid Cloud Thorugh FSx Cache

To address these challenges, Cadence, in collaboration with NetApp and AWS, has designed an exceptional architecture that simplifies data management in hybrid cloud bursting for demanding EDA workloads using Amazon FSx Netapp ONTAP (Amazon FSxN) FlexCache and Cadence Xcelium.

Cloud Bursting: A Quick Recap

With increasing design complexity and size, 3nm design projects require as much as 4-6X more computing and 4X more storage than prior processes, placing strain on the on-prem capacity. Further, the increasing adoption of advanced process nodes and generative AI and GPU-based EDA tools have triggered a significant surge in design data volumes.

Therefore, the underlying storage and data management system must be redesigned for market relevancy to support parallel cloud-scale workloads. The Cadence Cloud enables EDA design flows to scale compute with almost no limit on using the latest compute cores available. However, efficient resource utilization with the dynamic EDA demands is mandatory for keeping the costs low as no company has infinite resources (licenses and other infrastructure)

These factors are triggering quicker adoption of advanced silicon manufacturing technologies. To keep up, the semiconductor companies are either evaluating or are already using the cloud burst capacity.

With the advent of cloud bursting, new flexibility has arrived that designers can leverage when on-premises capacity is strained.

The Developer's Dilemma

Although cloud bursting offers numerous benefits, including unlimited resources and cost-efficiency, the challenges mentioned below make it difficult to use in the EDA industry.

• Data Management
  • Size of data.
  • Figuring out which Tools, libraries, and design data are needed to run in the cloud.
  • Data Mobility - How to transfer to the cloud.
  • Keeping the data in sync between the on-prem and cloud.
• Latency

Verification plays a vital role in chip design; verification is impossible to do exhaustively and is often the long pole in SoC design. Keeping data available, secure, and in sync with global design teams is a challenge.

With the globally spread teams working over the cloud, data management is a real pain. For instance, consider running full-chip verification, where 85% of the work can be accomplished on-premises, as the entire job would take longer than desired because of capacity limitations. The remaining work can be done in the cloud to speed up the completion process. Once the IT team has developed the hybrid solution, the engineer must perform the following tasks:

• Manually divide the design, identifying specific blocks for on-premises resources and others for the cloud.
• Transfer the relevant design data to the cloud.
• Retrieve the output data from the cloud and bring it back to the on-premises resources.
• Merge the cloud-generated and on-premises data in the on-premises infrastructure for analysis and processing for the next step in the flow.

This method can be challenging and time-consuming, causing significant overhead to engineering productivity. And the problem worsens when the designer must manage time, data, and cost!

How Cadence Resolves the Developer's Dilemma

Cadence, in collaboration with NetApp and AWS, to simplify data management, designed an architecture that allows customers to seamlessly burst into the cloud without modifying their workflow. The key features of the architecture are as follows:

• Burst to the cloud to get the performance/scale needed at peak times and debug the on-prem without changing workflows.
• FlexCache intelligently caches only the data your flow uses in the cloud.
• A complete solution for verification on AWS FSxN using Cadence Xcelium.

This "Design Anywhere" solution enables bursting by making tools and libraries appear local to the cloud. It no longer must uncover and track multiple layers of dependencies painstakingly and frequently replicate new versions to the cloud. Instead, everything you need in the cloud is pulled into a cache on demand, removing the IT burden of manual data management. Better still, just the files or parts of files are copied, as per the requirements.

The image above shows a sample Cadence Xcelium simulator workflow. Cadence Xcelium simulator, libraries, and inputs are quickly cached in the cloud when a verification job is initiated from the on-premises. The job then executes in the cloud using the cached files and writes the results to a scratch volume that seamlessly synchronizes with the on-premises storage volume. Engineers remain in their familiar on-premises environment, easily running jobs and accessing results without the need to transfer files to the cloud manually.

When a flow starts using a new library file, the file is read from the origin volume and cached in the cloud. The first read loads at the speed of the WAN connection. The subsequent reads sub-millisecond latency served from the cache. Also, it allows us to do the opposite and make results available on-prem by simply writing output to a scratch volume. Later, the data (only the files or parts of files as required) is automatically pulled down from the cloud if you need to debug the output on-prem. It embodies the perfect blend of scalability, performance, and security, addressing electronic design automation processes' unique needs.

Conclusion

Cadence collaboration with NetApp and AWS is a compelling solution to simplify data management in hybrid cloud bursting for demanding EDA workloads. The solution enables leveraging the right tools, and library versions become automatically available where you need them without replicating unnecessary data or debugging long dependency chains. This solution ensures optimal performance and scalability for EDA storage, which requires managing high metadata and parallelized access.

Join us for more details in NetApp's Cloud Track presentation at CadenceLIVE Silicon Valley on April 17, 2024.

Learn more 

• Cadence OnCloud solutions.
• Cadence Hybrid Cloud Simplifies Data Lift and Shift
• Book a meeting with Cadence, NetApp, and AWS to design with FSx for NetApp ONTAP.