Issue link: https://read.uberflip.com/i/1187300
Amazon Web Services – HPC Lens AWS Well-Architected Framework Page 5 solve your HPC problems. This will help you achieve greater impact for your computing solutions and datasets. • Cloud-native designs: It might not be necessary to replicate your on- premises environment when you move workloads to AWS. It is often suboptimal to do so. The breadth and depth of AWS services enables HPC workloads to run in new ways using new design patterns and cloud- native solutions. For example, each user or group can use a separate cluster, which can independently scale depending on the load. Users can rely on a managed service, like AWS Batch, or serverless AWS Lambda computing, to manage the underlying compute. Consider not using a traditional cluster scheduler. Instead, use a scheduler only if your workload requires it. In the cloud, clusters no longer require permanence. After you automate cluster deployment, you can tear one cluster down and launch a new one on demand, with different parameters. • Test real-world workloads: The only way to know how your production workload will perform on the cloud is to test it on the cloud. Most HPC applications are complex, and their memory, CPU, and network patterns often can't be reduced to a simple test. For this reason, generic benchmarks aren't reliable predictors of actual HPC production performance. Similarly, there is little value in testing an application with a small benchmark set or "toy problem." Since you will only pay for the hours of compute and storage you actually use, it's quite feasible to do a realistic proof-of-concept on AWS. This is a major advantage of a cloud- based platform: a realistic, full-scale test can be done before migration. • Balance time-to-results and cost reduction: Analyze performance using the most meaningful parameters: how long it will take, and how much it will cost. Cost optimization should be used for workloads that are not time-sensitive. On AWS, Spot Instances are frequently the least expensive method for such workloads. For example, if a researcher has a large number of lab measurements that need to be analyzed sometime before next year's conference, Spot Instances can help her analyze the largest possible number of measurements within her fixed research budget. For urgent workloads, such as emergency response modeling, cost optimization can be traded for performance, and instance type, procurement model, and cluster size should be chosen for the lowest execution time. If comparing between platforms, it's important to take