At Synopsys, we are constantly looking for ways to optimize the user experience of running workloads in the cloud by providing performance and cost optimization out of the box. Heterogeneous clusters and Dynamic scaling are the constructs designed to solve two fundamental problems:
- Which hardware is best suited for my workload
- How do I ensure cost efficiency running the workload in a performant fashion
One of the key benefits which Synopsys Cloud offers is providing our customers with a selection of optimized computing specific to each workload within the complex chip design lifecycle. We tested and benchmarked hundreds of compute virtual machine types available in the public cloud to enable a small subset optimized for specific EDA workloads. This selection of optimized compute enables customers to quickly establish the optimal compute, thereby reducing the wastage of resources and therefore cost optimization.
In a larger cluster, such groups of optimized hardware can be grouped together as sub-clusters of a heterogeneous cluster in an automated fashion. This allows for a standardized cluster that consists of optimized hardware for distinct EDA workloads with the operating system and EDA software along with it. Diverse EDA jobs can be submitted to the same cluster but directed to the most appropriate sub-clusters with further specifications of resource strings.