MLCOST04-BP02 Use managed build environments
Using managed build environments for machine learning development instead of local setups provides significant cost, time, and resource advantages. Managed notebooks come pre-configured with security, networking, storage, and compute capabilities that would otherwise require extensive development and maintenance effort. These environments also offer flexible machine selection, including access to powerful GPUs and high-memory instances that may be impractical in local setups.
Desired outcome: You can quickly start machine learning development work without spending time setting up infrastructure, managing dependencies, or configuring development environments. You gain access to scalable compute resources, including specialized hardware like GPUs, and benefit from built-in security and collaboration features, allowing you to focus on building ML models rather than managing infrastructure.
Common anti-patterns:
-
Spending excessive time configuring local development environments for each project.
-
Encountering hardware limitations when training complex models locally.
-
Struggling with inconsistent development environments across team members.
-
Managing security and networking configurations manually.
-
Inability to scale resources up or down based on workload requirements.
Benefits of establishing this best practice:
-
Reduced time to start development with pre-configured environments.
-
Access to powerful compute resources on demand.
-
Consistent development environments for team members.
-
Built-in security, networking, and storage capabilities.
-
Simplified collaboration and sharing of notebooks and models.
-
Cost optimization through pay-for-what-you-use model.
Level of risk exposed if this best practice is not established: Medium
Implementation guidance
When implementing machine learning projects, your development environment plays a critical role in productivity and efficiency. Local development environments often lead to inconsistencies between team members, dependency conflicts, and hardware limitations. Managed build environments address these challenges by providing standardized, scalable, and secure solutions for ML development.
Amazon SageMaker AI offers several managed environment options tailored to different user needs and expertise levels. These include SageMaker AI Notebook Instances for individual developers, SageMaker AI Studio for comprehensive ML development, and SageMaker AI Canvas for no-code ML solutions. These environments come pre-configured with the necessary tools and libraries, saving setup time and fostering consistency.
These managed environments integrate seamlessly with other AWS services, making it simple to access data stored in Amazon S3, use specialized hardware like GPUs, and deploy models to production endpoints. They also provide built-in security features, version control, and collaboration capabilities that would be difficult to implement in a local setup.
Implementation steps
-
Evaluate your ML development needs. Begin by assessing your team's requirements, including technical expertise, project complexity, and compute resource needs. Identify which SageMaker AI offering best matches these requirements.
-
Use Amazon SageMaker AI Notebook Instances. Set up SageMaker AI Notebook Instances which provide a fully managed Jupyter notebook environment. These instances come pre-loaded with popular ML frameworks and libraries, allowing you to start working immediately.
-
Implement Amazon SageMaker AI Studio. Deploy SageMaker AI Studio as your comprehensive ML development environment. SageMaker AI Studio provides a web-based visual interface where your team can perform ML development steps from data preparation to model deployment. Access Studio by creating a SageMaker AI domain through the SageMaker AI console, which enables team management and resource sharing capabilities.
-
Deploy SageMaker AI Canvas for business users. Implement SageMaker AI Canvas for business analysts and non-technical team members who need to create ML models without coding. Canvas provides an intuitive visual interface for importing data, creating models, and generating predictions.
-
Set up proper IAM roles and permissions. Configure appropriate IAM roles for your SageMaker AI environments to provide secure access to AWS resources. Create specific roles that follow the principle of least privilege, granting only the permissions necessary for your ML workflows.
-
Configure data access and storage. Set up connections between your SageMaker AI environments and data sources such as Amazon S3, Amazon Redshift, or Amazon RDS. Configure appropriate permissions to access these data sources securely.
-
Implement version control and collaboration. Integrate your managed environments with version control systems like Git to track changes to notebooks and code. Use SageMaker AI Studio's built-in collaboration features to share work among team members.
-
Optimize for cost efficiency. Configure auto-shutdown policies for notebook instances when they're idle to reduce costs. Monitor resource usage and adjust instance types as needed to balance performance and cost.
-
Use SageMaker AI HyperPod for large-scale training. For distributed training of large models, use SageMaker AI HyperPod which provides purpose-built infrastructure with automatic checkpoint storage and recovery, optimizing resource utilization for long-running training jobs.
-
Enable SageMaker AI JupyterLab 3 features. Take advantage of the productivity improvements in JupyterLab 3, which is available in both SageMaker AI Studio and Notebook Instances, providing better performance and enhanced features for developers.
Resources
Related documents:
