High Performance Computing (HPC)
As Nova’s approach relies in part on a model-based scattering physical solution, it necessitates finding the solution of the electromagnetic (Maxwell’s) equations to a generic scattering object whose geometry and materials properties span a wide range of values. Such solutions involve a heavy load of calculations with matrices as large as even ~1000X1000. These time & memory consuming operations include multiplications, inversions, eigen-value decompositions, Fourier transforms etc. This affects the time for creating and refining metrology solutions (called recipes) as part of the application development process.
To accelerate recipe preparation, the NovaMARS® software metrology modeling engine utilizes the Nova HPC® systems. Based on Microsoft Windows HPC Server platform, Nova HPC delivers scalability and fault tolerance. The solution combines a scalable, user-configurable infrastructure with Nova’s proprietary task management software designed to address the increasing complexity of IC manufacturing processes.
Highlights & Benefits
- Accelerate recipe setup: Using a parallel computing platform reduces the time it takes for new recipe library building.
- Scale-as-you-grow: The flexible, scalable infrastructure is designed to grow according to your business needs
- Cost-effective grid computing solution: Leveraging existing computation resources reduces the initial solution and expansion costs.
To decrease application development cycle time, NovaMARS is utilized for calculations of Nova High Performance Computing (HPC). The system is based on Microsoft Windows HPC server platform with scalability and fault tolerance ability. An end-user Grid Computing option is an alternative for increasing computing needs. Scalable and user configurable infrastructure with Nova’s proprietary task management software addresses the growing needs of IC manufacturing metrology.
WinHPC allows building libraries, running real-time tasks, and also supports real-time interpretation. WinHPC can scale the computation resources transparently inside the local grid and to external computing power on the Cloud. The system can support several grid subsystems connected together, and at the same time support several customers with a single system.
Basic architecture of an HPC Cluster
A cluster consists of several servers networked together where each server in the cluster performs one or more specific tasks. Cluster components include Head Nodes, and Compute Nodes, Job scheduler and Broker Nodes (for SOA enabled clusters).