AWS wellarchitected documentation change
Summary
Restructured design principles section with updated content including dynamic architectures, procurement models, data considerations, infrastructure as code, secure collaboration, cloud-native designs, workload testing, and cost/time balance. Added new subsections and rephrased introduction.
Security assessment
The 'Collaborate securely' bullet explicitly adds security guidance about cross-account data sharing with Amazon S3 and remote visualization with Amazon DCV. While this introduces security best practices, there is no indication it addresses a specific vulnerability or incident.
Diff
diff --git a/wellarchitected/latest/high-performance-computing-lens/design-principles.md b/wellarchitected/latest/high-performance-computing-lens/design-principles.md index c44a5c90e..f76105e11 100644 --- a//wellarchitected/latest/high-performance-computing-lens/design-principles.md +++ b//wellarchitected/latest/high-performance-computing-lens/design-principles.md @@ -3 +3 @@ -[Documentation](/index.html)[AWS Well-Architected](https://aws.amazon.com/architecture/well-architected/)[AWS Well-Architected Framework](welcome.html) +[Documentation](/index.html)[AWS Well-Architected](https://aws.amazon.com/architecture/well-architected/)[AWS Well-Architected Framework](high-performance-computing-lens.html) @@ -5 +5 @@ -# Design Principles +# Design principles @@ -7 +7 @@ -In the cloud, a number of principles drive operational excellence. In particular, the following are emphasized for HPC workloads. See also the design principles in the [AWS Well-Architected Framework whitepaper](https://docs.aws.amazon.com/wellarchitected/latest/framework/welcome.html). +Consider the following design principles when building an HPC workload in the cloud: @@ -9 +9 @@ In the cloud, a number of principles drive operational excellence. In particular - * **Automate cluster operations** : In the cloud, you can define your entire workload as code and update it with code. This enables you to automate repetitive processes or procedures. You benefit from being able to consistently reproduce infrastructure and implement operational procedures. This includes automating the job submission process and responses to events, such as job start, completion, or failure. In HPC, it is common for users to expect to repeat multiple steps for every job including, for example, uploading case files, submitting a job to a scheduler, and moving result files. Automate these repetitive steps with scripts or by event-driven code to maximize usability and minimize costs and failures. + * **Use dynamic architectures** : Avoid defaulting to static architectures and cost estimates that use a steady-state model. Your architecture can be dynamic, growing and shrinking to match your HPC demands over time. Match your architecture design and cost management explicitly to the natural cycles of HPC activity. @@ -11 +11,13 @@ In the cloud, a number of principles drive operational excellence. In particular - * **Use cloud-native architectures where applicable** : HPC architectures typically take one of two forms. The first is a traditional cluster configuration with a login instance, compute nodes, and job scheduler. The second a cloud-native architecture with automated deployments and managed services. A single workload can run for each (ephemeral) cluster or use serverless capabilities. Cloud-native architectures can optimize operations with democratizing advanced technologies; however, the best technology approach aligns with the desired environment for HPC users. + * **Align the procurement model to the workload** : AWS makes a range of compute procurement models, such as on-demand and savings plans, available for different HPC demand patterns. By selecting the correct model, you only pay for what you need. You can potentially combine models for a committed rate while having burst capacity with on-demand compute. + + * **Consider your data** : Understand your data before you begin designing your architecture. Consider your data's location, size, update, and regulatory requirements. A holistic optimization of performance and cost focuses on compute and includes data considerations. Your data requirements may shape optimal performance based on available resources in your selected Regions. + + * **Automate with infrastructure as code:** Consider products, such as AWS ParallelCluster and AWS Research and Engineering Studio, to ease your effort in automating your infrastructure. Alternatively, you can customize your infrastructure with services, such as AWS CloudFormation, if needed but plan for a higher development effort. + + * **Collaborate securely** : HPC work often occurs in a collaborative context and is usually part of a larger workflow. Take full advantage of the security and collaboration features that make AWS an excellent environment for you and your collaborators to solve your HPC problems. For example, you can grant cross-account access to share a full or partial data set with Amazon S3 to another AWS account. This helps your computing solutions and datasets achieve a greater impact by securely sharing within a selective group or publicly sharing with the broader community. When collaborating, consider data locality and workflow architecture. For example, use remote visualization with Amazon DCV rather than transferring data to users. + + * **Use designs built for the cloud** : Directly replicating your on-premises architecture is usually unnecessary and suboptimal when migrating workloads to AWS. Consider the breadth and depth of AWS services to create new design patterns with solutions that are built for cloud architectures _._ For example, in the cloud, each user or group can use a separate cluster, which can independently scale depending on the load. + + * **Test real-world workloads** : You only pay for what you actually use with AWS, which makes it possible to create a realistic proof-of-concept with your own representative models. Most HPC applications are complex, and their memory, CPU, and network patterns often can't be reduced to simplified microbenchmarks or theoretical performance benchmarks. With AWS, you can quickly get started, iterate, and optimize your design with only paying for consumed resources as you finalize your architecture. + + * **Balance your desired price and time-to-results** : Use time and cost to analyze performance. Sometimes you can prioritize reducing cost when you do not immediately require results. For time-critical workloads, consider trading cost optimization for faster time-to-results. Determine your desired price-for-performance approach based on your results timeline and available budget. @@ -22 +34 @@ To use the Amazon Web Services Documentation, Javascript must be enabled. Please -Operational Excellence +Abstract and introduction @@ -24 +36 @@ Operational Excellence -Best Practices +Definitions