Cloud computing in electronic design automation (EDA) can improve productivity and expedite time to market. Running both Calibre® early design stage and signoff verification in the cloud is a perfect combination for design companies who are designing large, leading-edge chips in advanced nanometer processes. Calibre flows not only enable fast design turns using modest compute resources, but also consistently get faster release after release. To take full advantage of the benefit of running Calibre in a cloud environment, integrated circuit (IC) design teams should understand the structure and value of the different options provided by their cloud provider, and optimize their cloud configurations and processes to best suit their design needs, business requirements, and computing budgets. To help design companies in this process, Calibre Design Solutions initiated a collaboration between Amazon Web Services (AWS), Amazon Annapurna Labs (Annapurna) and Siemens EDA to run and evaluate results from a series of Calibre flow benchmark runs in a cloud environment. Multiple Calibre flows were exercised on two full-chip production designs provided by Annapurna. Based on those results and observations, we established a set of best known methods (BKMs) that design companies and design teams can use as a roadmap when creating their optimal cloud computing environments.
Siemens EDA collaborated with AWS and Annapurna Labs to evaluate results from a series of Calibre flow benchmarks run inside Annapurna’s production virtual private cloud (VPC), which is hosted on AWS. Running Calibre applications in the cloud can provide a twofold advantage: design teams don’t have to wait for local compute resources to become available, and design companies have the option to increase available compute resources during peak or critical demand periods, or when they want to take advantage of Calibre scalability to boost the number of design turns per day. Running multiple Calibre applications in a variety of scenarios and conditions enabled us to define a set of best-known methods (BKMs) for running Calibre in the cloud. These BKMs include general recommendations for EDA in the cloud, as well as specific recommendations for optimizing Calibre IC design verification in a cloud environment.