Enable Semiconductor Design on AWS with Intervision’s DesignHub

Overview

Electronic Design Automation (EDA) workloads present unique challenges to compute grids, complex file-based storage systems, and scalability requirements. Semiconductor companies are rapidly growing compute and storage resources needed to meet their product complexity and density, development lifecycle and time to market requirements. Semiconductor clients are leveraging virtually unlimited compute, storage, and other resources available on AWS, while smaller startups and universities lack this ability.

In this blog post, we will talk about the challenges in EDA workload development and how Intervision’s DesignHub can help clients easily deploy and operate a multi-user environment for computationally intensive workflows such as Computer-Aided Engineering (CAE). The solution features a large selection of compute resources, a fast network backbone, virtually unlimited storage, and budget & cost management directly integrated within AWS for anyone that needs security for intellectual property (IP) and Computer-Aided Designs (CAD). This solution deploys a user interface (UI) with cloud workstations, file management, and automation tools that enable you to create your own queues, scheduler resources, Amazon Machine Images (AMIs), and management functions for user and group permissions.

Challenges in EDA workload development

Productivity and time-to-market

Productivity and time-to-market are the keys for profitability and a successful design. Semiconductor startup and universities have limited time and budgets to bring a chip to production-ready designs. Startups that are fine-tuned but three months late to market tend to lose 30% of their profits; in contrast, if they are on time but 50% over budget, they may only lose 4% of their profits. In short, getting to market earlier equates to greater profitability in the long run, even if the project is over budget at time of launch. Universities operate on a semester cycle that requires immediate availability to resources. Our targeted client prefers to spend funds on design resources and the actual compute they use when they need it, rather than spend money on compute infrastructure that depreciates over 3-5 years. By leveraging the elasticity of the cloud, startups and universities can maximize the cash they have to get to market.

Design complexity

Advanced chip design requires not only the latest EDA Tools/IP, but also requires a large and constantly growing compute infrastructure. Compute demands nearly double with each process node (chip size and density). Obtaining, provisioning, housing, and maintaining the infrastructure required for advanced chip design is an enormous challenge for the semiconductor design value chain (fabs, EDA tool vendors, and chip designers).

Security challenges

Semiconductor design chains are challenged to provide sufficient IT and IP security. From foundry process design kits management process to tape-out file transfer, to third-party IP security requirements, the design environment must meet all these contractual obligations. Startups, universities, and fabs prefer solutions that meet their internal security requirements while achieving time-to-value.

IT management resources

Designers often start and manage semiconductor startups and universities specialized in CAE design and have few dedicated IT resources. Design engineers often handle IT system setup and management, which can distract them from actual design work.

How can DesignHub help semiconductor clients?

DesignHub deploys a scalable, high performance and secure design and verification environment that serves relieves the burden of managing a cloud environment for the clients who have limiting understanding of the cloud, and ensures that the clients cannot leak FAB/EDA intellectual property (IP) technology. InterVision has further refined the environment to be cost neutral to on-premises environments with scalable features to ensure price points are competitive for developers of complex and intensive designs.

DesignHub defines and automatically configures an optimized environment for semiconductor design, supporting EDA and related IP from third-party partners including Synopsys, Cadence, Siemens/Mentor, Ansys, and Arm. The environment enables both batch and interactive workflows, as well as enabling secure collaboration between globally distributed semiconductor design and verification teams, their intellectual property (IP) partners, fabs, and EDA tool vendors. The environment also allows semiconductor design and verification teams to augment their EDA workflows using more advanced, cloud-native services, such as machine learning for higher quality-of-results, by offering chip design all the way to GDSII (tape-out) and allowing customers to bring their own EDA licenses to the DesignHub.

Bring your own licenses

DesignHub automates the entire license server standup process, including FLEXlm license use manager software for EDA Tool and IP Partners. InterVision has integrated the solution to enable multi-vendor tool usage for chip design in one centrally managed location with provisioned tools and flexibility of customization. Furthermore, DesignHub integrates a license use manager backup to align with client requirements. As a result, semiconductor clients’ current license agreement and tool provider relationship is unaffected.

Centrally manage your semiconductor design environment

DesignHub centralizes IT responsibilities for clients whether our client’s design teams are regional or global. Leveraging AWS availability zones with an understanding of tool latency requirements, DesignHub brings EDA tools to our client’s engineering teams versus deploying HPC systems remotely. InterVision’s IV Flex solution provides clients a managed service for their design environment, EDA Tools/IP, and user onboarding. Consequently, DesignHub allows our clients’ design teams to focus on the actual work of chip development.

EDA job scheduler integration and optimization

DesignHub can deploy both commercially available schedulers (LSF, SGE, AltAir Accelerator, etc.) as well as the opensource version (SLURM, OpenPBS, etc.). InterVision’s workflow consultants can advise clients on selecting the right scheduler solution to fit their short-term and long-term needs. Clients that are operating both on on-premises and in cloud environments can take advantage of DesignHub’s hybrid mode with pre-built job scheduler automation and integration. InterVision’s workflow engineers specialize in LSF and LSF multi-cluster deployments.

EDA cost optimization

DesignHub integrates Spot Instances for EDA workload cost optimization. In addition to cost optimization, DesignHub allows clients to burst workloads using hyper-scaling to optimize the compute resources requirements, which decreases critical license use cost and job runtimes.

Secure IP chamber

DesignHub incorporates AWS-approved secure chambers with advanced IP security controls built in, including monitoring and alerts enabling designers to remove infrastructure barriers. DesignHub protects client IP at a desktop and system level using AWS’s virtual desktop and security features in AppStream and Workspaces, providing a fully secure design environment. Furthermore, DesignHub implements the AWS secure collaboration reference architecture and is built to mitigate possibilities of design artifacts, IP, and design file exfiltration.

DesignHub’s Implementation Overview

DesignHub is based on AWS Solution Scale-Out Computing (SOCA), which deploys a cluster and provides automated cluster provisioning orchestration. In addition, it features a large selection of compute resources to include a fast network backbone, nearly unlimited storage, and budget & cost management directly integrated with AWS. AWS services within scope include Amazon Elastic Compute Cloud (EC2 Spot and EC2 On-Demand Instances), Amazon Simple Storage Service (S3), and FXs for NetApp ONTAP, and FLEXlm license use manager.

In its POC stage, DesignHub had a requirement to create a secure cloud-based EDA design environment, accessible from on-premises office or university, leveraging software tools from Ansys, Cadence, Siemens (MG) and Synopsys and for an initial design team of five to ten users. The solution needed to support remote users to securely connect to the environment’s tools from remote locations, due to COVID-19 restrictions and work-from-home requirements.

AWS and InterVision proposed a solution that leverages the AWS Virtual Private Cloud (VPC) and Elastic Compute Cloud (EC2) services to create this secure, private environment. This solution utilizes the AWS Site-to-Site VPN service to create an encrypted private connection between an office network firewall/router and the AWS Virtual Private Cloud. As a result, this solution allowed users on-premises at office Local Area Network (LAN) to connect to the License Servers, File Shares, and other AWS resources without additional steps.

The DesignHub solution was deployed into the dedicated “RapidStart” account associated with a foundry customer of AWS. This organization will leverage the AWS Control Tower service to ensure that all associated AWS accounts follow AWS best practices. For example, AWS backup rules will be created at the organization-level to backup resources based on tags.

Amazon for NetApp ONTAP will be leveraged to provide encrypted file storage and file shares accessible from on-premises and AWS Linux, Windows, and MacOS clients.

An Active Directory server was deployed for directory services, user accounts, and access to the FSx file shares. A self-managed Active Directory was created and configured to provide these services during an initial engagement; ongoing operations are out of scope and can be provided separately via InterVison’s IV Flex managed services.

InterVision recommends investigating the suitability of the AWS AppStream and WorkSpaces services to serve the modeling workstations for users both on the on-premises network, as well as for remote users in the classroom, at home or on the road. These services provide the ability to easily scale the number of workstations to accommodate user demand, eliminating resource contention. The overall security posture has significantly improved as well, since no applications or data are stored on the user’s local or personal machines: data remains on infrastructure controlled by DesignHub and the client at all times.

Leveraging the AppStream or WorkSpaces services will eliminate the need to provide a VPN connection for external remote users, as they would have encrypted access to their desktop applications through a secure, web-based portal. If some remote users do require VPN access directly to the DesignHub environment from their local computer, the AWS Client VPN service can provide this functionality.

On-demand EC2 instances were created for the Cadence, Ansys/MatLab, and Synopsys software license servers in the AWS environment. Site-to-Site VPN has been configured between the POC on-premises office and the AWS environment, allowing for the license servers to be accessible from both AWS and the client’s offices.

To facilitate scaling the number of modeling and simulation EC2 instances and modeling AppStream/WorkSpaces instances, InterVision will work with each client to produce “golden” images with all required software packages installed. These Amazon Machine Images (AMIs) will be used to deploy the Simulation EC2 instances, and the AppStream or WorkSpaces images will be used with the respective services to provide the streaming user workstations.

InterVision will deploy and configure into this environment the following resources:

AWS Control Tower/AWS Organizations setup: Management, Audit and Logging account creation & configuration, and Workload account
VPC, Subnet, Route Table, Internet Gateway, NAT Gateway, and related resource creation
1 Self-Managed Active Directory (Windows 2019 Server, t3a.large)
1 Amazon FSx for NetApp ONTAP shared file system (2 TB initial size)
1 Ansys/MatLab License Server (CentOS Linux, t3a.large)
1 Ansys/MatLab Modeling EC2 Instance (from AMI) or AppStream/WorkSpace (from Image)
1 Ansys/MatLab “small simulation” EC2 Instance (from AMI)
1 Ansys/MatLab “large simulation” EC2 Instance (from AMI)
1 Ansys/Matlab Workstation/Simulation AMI and 1 AppStream or WorkSpaces Image
1 Cadence License Server (CentOS Linux, t3a.large)
1 Cadence Workstation/Simulation AMI and 1 AppStream or WorkSpaces Image
1 Synopsys License Server (CentOS Linux, t3a.large)
1 Synopsys Workstation/Simulation AMI and 1 AppStream or WorkSpaces Image
1 SolidWorks License/PDM Vault Server (Windows 2019 Server w/SQL Server 2017 Standard, t3a.xlarge)
1 SolidWorks Workstation/Simulation AMI and 1 AppStream or WorkSpaces Image
1 SolidWorks “modeling” EC2 Instance (from AMI) or AppStream/WorkSpaces (from Image)
1 SolidWorks “small simulation” EC2 Instance (from AMI)
Up to 5 additional (10 total) EC2 instances from AMIs

Initially, startup and university clients will desire to rapidly deploy the core infrastructure necessary to provision EC2 License Servers for these software tools and begin the procurement process for license keys from these vendors. As such, the AWS account provisioning, initial VPC/Subnet configuration, and License Server EC2 Instance creation activities have been prioritized.

Upon the initial project kick-off, InterVision will review the DesignHub solution with the client and gather prerequisite information for creating the AWS Organization Account and related sub-accounts, deploying a standard for the DesignHub VPC resources, such as IP/CIDR block assignments and any other necessary information.

InterVision will deliver consulting and professional services according to these priorities together with identified task dependencies and resource availability. InterVision will staff the project according to a fixed weekly schedule determined by the staffing plan. When delays and waits occur, InterVision resources will maintain forward progress of the overall project by directing efforts to other in-scope tasks. Additional resources are available via change request on an hourly basis.

Conclusion

In this blog post, we discussed the challenges in EDA workload development and how Intervision’s DesignHub can help clients easily deploy and operate a multi-user environment for computationally intensive workflows such as Computer-Aided Engineering (CAE). DesignHub features a large selection of compute resources, a fast network backbone, virtually unlimited storage, and budget & cost management directly integrated within AWS for anyone that needs security for intellectual property (IP) and Computer-Aided Designs (CAD). Contact us for a demo or a deeper dive on DesignHub using the following link.

Related Resources:

Relevant Resources:

AWS for Industries