Taking EDA To The Cloud

By now, virtually everyone knows about the “cloud”—that amorphous delivery of computing services over the Internet. Servers, storage, databases, networking, software, analytics, intelligence, and more are all on offer. Clouds may be private and limited to a single organization (enterprise clouds), be available to many organizations (public cloud), or a combination of both (hybrid cloud).

The most common cloud service applications are:

• Infrastructure as a Service (IaaS), where infrastructure is hosted by the cloud provider and supplied to various users,
• Platform as a Service (PaaS), which provides a platform that allows users to develop, run, and manage applications without the complexity of putting together and maintaining an on-site grid,
• Software as a Service (SaaS), an “on-demand” software licensing and delivery model that enables users to access software without having to have it installed on their hardware.

The appeal is obvious—companies typically pay only for cloud services they use, helping to lower operating costs, run infrastructure more efficiently, and scale quickly as business needs change. Taking advantage of cloud computing can also lead to faster innovation, more flexible resource usage, and always-desirable economies of scale.

Cloud computing for EDA
It took some time for the semiconductor and electronic design automation (EDA) industries to become comfortable with the cloud services model, primarily because of their concerns over the protection and control of intellectual property (IP) and highly proprietary, highly sensitive data such as process design kits (PDKs). Cloud providers responded to these concerns by enhancing and extending security measures, and the IP providers (most notably the foundries) returned the favor by adopting the use of cloud technology. The PaaS model is the most common form of cloud technology in the semiconductor industry, in which EDA companies enable their customers to develop and run their PDKs on a cloud platform.

In fact, running EDA tools on the cloud doesn’t require any special processes on the part of the user. Any EDA software that can run locally can be installed and run on the cloud. The two leading reasons design companies and foundries use cloud services are costs and latencies. From a cost perspective, using PaaS to run EDA software on selected designs allows companies to avoid the capital investment of buying servers to provide the necessary processing power and runtime, as well as the ongoing cost of maintaining and supporting those resources. For companies who regularly experience high demand periods in which multiple users need to access their server’s limited resources, PaaS can provide transitions between low and high demand intervals that are invisible to the end users.

With 2x more transistors being added node over node, and ever-larger and more complex foundry rule decks, the amount of computing required for today’s integrated circuit (IC) designs makes it challenging, if not impossible, for most companies to obtain, install, and support the requisite amount of computing resources needed to maintain fast verification runtimes while ensuring design quality (Figure 1).

Figure 1: Industry trends for physical verification requirements per node.

Getting the most from cloud computing
Cloud computing provides companies an opportunity to accelerate the time to market for designs, particularly when you consider the growth in computing requirements at advanced nodes. However, as advantageous as hosting software on the cloud can sound, it actually does not make sense for all jobs. Companies need to do their analysis up front to understand why, how, and when they might benefit from using cloud computing instead of adding local resources.

Using the Calibre nmPlatform as our EDA toolsuite, we’ll look at how a company can utilize cloud computing to their benefit during design verification. Obviously, your mileage may vary, depending on what verification tools you use.

As previously mentioned, one of the reasons for using a PaaS model is to avoid the cost of acquiring and maintaining hardware. Increasing the number of CPUs on-site is never a simple task—hardware acquisition, grid installation, and maintenance all consume time and money. More importantly, on-site grids lack immediate additional resource availability. The good news is that the effort required to perform physical verification in the cloud is solely in the cloud setup itself— choosing the cloud provider and setting up the cloud environment.

Since the first introduction of Calibre hyper-remote capability in 2006, Mentor has worked constantly to help semiconductor companies and foundries achieve the greatest efficiency and value from their hardware resources by enabling the Calibre architecture to support scaling to large numbers of CPUs/cores (Figure 2).

Figure 2: Calibre nmDRC runtime vs. number of CPUs. (source: AMD. Used by permission)

Now, by removing the cost and latency barriers to resource usage, cloud computing allows companies to get instant access to the CPU resources they need, while still leveraging the platform’s inherent scaling capability, enabling them to efficiently achieve their physical verification goals (e.g., overnight runtimes) even in the face of exponential compute growth at the newest technology nodes.

Having access to EDA technology in the cloud can also provide a fast, cost-effective means of dealing with late-breaking emergencies. If all of a company’s internal resources are committed, but a critical issue requiring immediate resolution arises, cloud computing provides the means to resolve the emergency without causing major disruption anywhere else.

Making the cloud more cost-effective
Of course, cloud resources aren’t free. In the PaaS model, most users pay for cloud computing based on how much time they use. To ensure users can employ cloud resources in the most cost-efficient manner, we recommend certain usage guidelines and suggest best practices. While many of these suggestions apply to both in-house resources or cloud computing, adopting these practices helps ensure that companies can effectively manage their cloud computing costs.

Foundry rule decks: First and foremost is the recommendation to use the most recent foundry-qualified rule deck. Doing so ensures that the most recent coding best practices are adopted. In addition, because the Calibre engine is optimized for every release, users are assured of optimized runtimes and memory consumption, as shown in Figure 3.

Figure 3: (left) Normalized memory vs. Calibre release versions, (right) Normalized runtime vs. Calibre release versions.

Hierarchical filing: Implementing a hierarchical filing methodology, in which a design is sorted into cells that are later referred to in the top levels of the design, significantly reduces data size and enables a significant reduction in final sign-off runtimes.

Minimizing idle resources: Operations that require many resources often acquire all of those resources at the beginning of execution, even though most of those resources will be idle for significant periods of time. By implementing processes and technology that connect to resources only as they are needed, users can ensure they are minimizing their total cloud costs by reducing or eliminating idle resources.

Cloud workflow efficiency: By choosing geographically close cloud servers, you can reduce network latency time. Cache-based systems will also improve machine performance.

To minimize upload time, upload each block separately as it is available, along with standard cells and IPs, then upload the routing. By uploading in stages, you avoid any bottlenecks. You can then use the Calibre DESIGNrev interface in the cloud to assemble all the data (Figure 4).

Figure 4: Uploading blocks and routing separately, and combining the data in the cloud server, minimizes both upload time and potential bottlenecks.

Conclusion
Cloud processing provides companies an opportunity to reduce time to market and speed up innovation while maintaining or lowering operating costs. EDA technology is generally cloud-ready; with improvements in cloud security eliminating the industry concern over IP protection, the only barrier hindering the implementation of EDA technology in the cloud processing model has been removed.

Best practices enable companies to achieve maximum benefit from their transition to a cloud processing model, and to effectively manage their cloud computing costs. As IC companies increasingly look to leverage cloud capacity for faster turnaround times on advanced process node designs, they can be confident that running EDA in the cloud will provide the same sign-off verification results they know and trust, while enabling them to adjust their resource usage to best fit their business requirements and market demands.

For more information, read our new whitepaper “Calibre in the cloud: Unlocking massive scaling and cost efficiencies.”