# SUS 5 How do you select and use cloud hardware and services in your architecture to support your sustainability goals?
Look for opportunities to reduce workload sustainability impacts by making changes to your hardware management practices. Minimize the amount of hardware needed to provision and deploy, and select the most efficient hardware and services for your individual workload.
**Topics**
+ [SUS05-BP01 Use the minimum amount of hardware to meet your needs](sus_sus_hardware_a2.md)
+ [SUS05-BP02 Use instance types with the least impact](sus_sus_hardware_a3.md)
+ [SUS05-BP03 Use managed services](sus_sus_hardware_a4.md)
+ [SUS05-BP04 Optimize your use of hardware-based compute accelerators](sus_sus_hardware_a5.md)
# SUS05-BP01 Use the minimum amount of hardware to meet your needs
Use the minimum amount of hardware for your workload to efficiently meet your business needs.
**Common anti-patterns:**
+ You do not monitor resource utilization.
+ You have resources with a low utilization level in your architecture.
+ You do not review the utilization of static hardware to determine if it should be resized.
+ You do not set hardware utilization goals for your compute infrastructure based on business KPIs.
**Benefits of establishing this best practice:** Rightsizing your cloud resources helps to reduce a workload’s environmental impact, save money, and maintain performance benchmarks.
**Level of risk exposed if this best practice is not established:** Medium
## Implementation guidance
Optimally select the total number of hardware required for your workload to improve its overall efficiency. The AWS Cloud provides the flexibility to expand or reduce the number of resources dynamically through a variety of mechanisms, such as [AWS Auto Scaling](https://aws.amazon.com/autoscaling/), and meet changes in demand. It also provides [APIs and SDKs](https://aws.amazon.com/developer/tools/) that allow resources to be modified with minimal effort. Use these capabilities to make frequent changes to your workload implementations. Additionally, use rightsizing guidelines from AWS tools to efficiently operate your cloud resource and meet your business needs.
**Implementation steps**
+ Choose the instances type to best fit your needs.
+ [ How do I choose the appropriate Amazon EC2 instance type for my workload? ](https://aws.amazon.com/premiumsupport/knowledge-center/ec2-instance-choose-type-for-workload/)
+ [ Attribute-based instance type selection for Amazon EC2 Fleet. ](https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/ec2-fleet-attribute-based-instance-type-selection.html)
+ [ Create an Auto Scaling group using attribute-based instance type selection. ](https://docs.aws.amazon.com/autoscaling/ec2/userguide/create-asg-instance-type-requirements.html)
+ Scale using small increments for variable workloads.
+ Use multiple compute purchase options in order to balance instance flexibility, scalability, and cost savings.
+ [On-Demand Instances](https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/ec2-on-demand-instances.html) are best suited for new, stateful, and spiky workloads which can’t be instance type, location, or time flexible.
+ [Spot Instances](https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/using-spot-instances.html) are a great way to supplement the other options for applications that are fault tolerant and flexible.
+ Leverage [Compute Savings Plans](https://aws.amazon.com/savingsplans/compute-pricing/) for steady state workloads that allow flexibility if your needs (like AZ, Region, instance families, or instance types) change.
+ Use instance and availability zone diversity to maximize application availability and take advantage of excess capacity when possible.
+ Use the rightsizing recommendations from AWS tools to make adjustments on your workload.
+ [AWS Compute Optimizer](https://aws.amazon.com/compute-optimizer/)
+ [AWS Trusted Advisor](https://aws.amazon.com/premiumsupport/technology/trusted-advisor/)
+ Negotiate service-level agreements (SLAs) that allow for a temporary reduction in capacity while automation deploys replacement resources.
## Resources
**Related documents:**
+ [ Optimizing your AWS Infrastructure for Sustainability, Part I: Compute ](https://aws.amazon.com/blogs/architecture/optimizing-your-aws-infrastructure-for-sustainability-part-i-compute/)
+ [ Attirbute based Instance Type Selection for Auto Scaling for Amazon EC2 Fleet ](https://aws.amazon.com/blogs/aws/new-attribute-based-instance-type-selection-for-ec2-auto-scaling-and-ec2-fleet/)
+ [AWS Compute Optimizer Documentation ](https://docs.aws.amazon.com/compute-optimizer/index.html)
+ [Operating Lambda: Performance optimization](https://aws.amazon.com/blogs/compute/operating-lambda-performance-optimization-part-2/)
+ [Auto Scaling Documentation](https://docs.aws.amazon.com/autoscaling/index.html)
**Related videos:**
+ [ Build a cost-, energy-, and resource-efficient compute environment ](https://www.youtube.com/watch?v=8zsC5e1eLCg)
**Related examples:**
+ [ Well-Architected Lab - Rightsizing with AWS Compute Optimizer and Memory Utilization Enabled (Level 200) ](https://www.wellarchitectedlabs.com/cost/200_labs/200_aws_resource_optimization/5_ec2_computer_opt/)
# SUS05-BP02 Use instance types with the least impact
Continually monitor and use new instance types to take advantage of energy efficiency improvements.
**Common anti-patterns:**
+ You are only using one family of instances.
+ You are only using x86 instances.
+ You specify one instance type in your Amazon EC2 Auto Scaling configuration.
+ You use AWS instances in a manner that they were not designed for (for example, you use compute-optimized instances for a memory-intensive workload).
+ You do not evaluate new instance types regularly.
+ You do not check recommendations from AWS rightsizing tools such as [AWS Compute Optimizer.](https://aws.amazon.com/compute-optimizer/)
**Benefits of establishing this best practice:** By using energy-efficient and right-sized instances, you are able to greatly reduce the environmental impact and cost of your workload.
**Level of risk exposed if this best practice is not established:** Medium
## Implementation guidance
Using efficient instances in cloud workload is crucial for lower resource usage and cost-effectiveness. Continually monitor the release of new instance types and take advantage of energy efficiency improvements, including those instance types designed to support specific workloads such as machine learning training and inference, and video transcoding.
## Implementation steps
+ Learn and explore instance types which can lower your workload environmental impact.
+ Subscribe to [What's New with AWS](https://aws.amazon.com/new/) to be up-to-date with the latest AWS technologies and instances.
+ Learn about different AWS instance types.
+ Learn about AWS Graviton-based instances which offer the best performance per watt of energy use in Amazon EC2 by watching [re:Invent 2020 - Deep dive on AWS Graviton2 processor-powered Amazon EC2 instances](https://www.youtube.com/watch?v=NLysl0QvqXU) and [Deep dive into AWS Graviton3 and Amazon EC2 C7g instances](https://www.youtube.com/watch?v=WDKwwFQKfSI&ab_channel=AWSEvents).
+ Plan and transition your workload to instance types with the least impact.
+ Define a process to evaluate new features or instances for your workload. Take advantage of agility in the cloud to quickly test how new instance types can improve your workload environmental sustainability. Use proxy metrics to measure how many resources it takes you to complete a unit of work.
+ If possible, modify your workload to work with different numbers of vCPUs and different amounts of memory to maximize your choice of instance type.
+ Consider transitioning your workload to Graviton-based instances to improve the performance efficiency of your workload.
+ [AWS Graviton Fast Start](https://aws.amazon.com/ec2/graviton/fast-start/)
+ [Considerations when transitioning workloads to AWS Graviton-based Amazon Elastic Compute Cloud instances](https://github.com/aws/aws-graviton-getting-started/blob/main/transition-guide.md)
+ [AWS Graviton2 for ISVs](https://docs.aws.amazon.com/whitepapers/latest/aws-graviton2-for-isv/welcome.html)
+ Consider selecting the AWS Graviton option in your usage of [AWS managed services.](https://github.com/aws/aws-graviton-getting-started/blob/main/managed_services.md)
+ Migrate your workload to Regions that offer instances with the least sustainability impact and still meet your business requirements.
+ For machine learning workloads, take advantage of purpose-built hardware that is specific to your workload such as [AWS Trainium](https://aws.amazon.com/machine-learning/trainium/), [AWS Inferentia](https://aws.amazon.com/machine-learning/inferentia/), and [Amazon EC2 DL1.](https://aws.amazon.com/ec2/instance-types/dl1/) AWS Inferentia instances such as Inf2 instances offer up to 50% better performance per watt over comparable Amazon EC2 instances.
+ Use [Amazon SageMaker AI Inference Recommender](https://docs.aws.amazon.com/sagemaker/latest/dg/inference-recommender.html) to right size ML inference endpoint.
+ For spiky workloads (workloads with infrequent requirements for additional capacity), use [burstable performance instances.](https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/burstable-performance-instances.html)
+ For stateless and fault-tolerant workloads, use [Amazon EC2 Spot Instances](https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/using-spot-instances.html) to increase overall utilization of the cloud, and reduce the sustainability impact of unused resources.
+ Operate and optimize your workload instance.
+ For ephemeral workloads, evaluate [instance Amazon CloudWatch metrics](https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/viewing_metrics_with_cloudwatch.html#ec2-cloudwatch-metrics) such as `CPUUtilization` to identify if the instance is idle or under-utilized.
+ For stable workloads, check AWS rightsizing tools such as [AWS Compute Optimizer](https://aws.amazon.com/compute-optimizer/) at regular intervals to identify opportunities to optimize and right-size the instances.
+ [ Well-Architected Lab - Rightsizing Recommendations ](https://wellarchitectedlabs.com/cost/100_labs/100_aws_resource_optimization/)
+ [ Well-Architected Lab - Rightsizing with Compute Optimizer ](https://wellarchitectedlabs.com/cost/200_labs/200_aws_resource_optimization/)
+ [ Well-Architected Lab - Optimize Hardware Patterns and Observice Sustainability KPIs ](https://wellarchitectedlabs.com/sustainability/200_labs/200_optimize_hardware_patterns_observe_sustainability_kpis/)
## Resources
**Related documents:**
+ [Optimizing your AWS Infrastructure for Sustainability, Part I: Compute](https://aws.amazon.com/blogs/architecture/optimizing-your-aws-infrastructure-for-sustainability-part-i-compute/)
+ [AWS Graviton](https://aws.amazon.com/ec2/graviton/)
+ [Amazon EC2 DL1](https://aws.amazon.com/ec2/instance-types/dl1/)
+ [Amazon EC2 Capacity Reservation Fleets](https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/cr-fleets.html)
+ [Amazon EC2 Spot Fleet](https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/spot-fleet.html)
+ [Functions: Lambda Function Configuration](https://docs.aws.amazon.com/lambda/latest/dg/best-practices.html#function-configuration)
+ [ Attribute-based instance type selection for Amazon EC2 Fleet ](https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/ec2-fleet-attribute-based-instance-type-selection.html)
+ [Building Sustainable, Efficient, and Cost-Optimized Applications on AWS](https://aws.amazon.com/blogs/compute/building-sustainable-efficient-and-cost-optimized-applications-on-aws/)
+ [ How the Contino Sustainability Dashboard Helps Customers Optimize Their Carbon Footprint ](https://aws.amazon.com/blogs/apn/how-the-contino-sustainability-dashboard-helps-customers-optimize-their-carbon-footprint/)
**Related videos:**
+ [Deep dive on AWS Graviton2 processer-powered Amazon EC2 instances](https://www.youtube.com/watch?v=NLysl0QvqXU)
+ [Deep dive into AWS Graviton3 and Amazon EC2 C7g instances](https://www.youtube.com/watch?v=WDKwwFQKfSI&ab_channel=AWSEvents)
+ [ Build a cost-, energy-, and resource-efficient compute environment ](https://www.youtube.com/watch?v=8zsC5e1eLCg)
**Related examples:**
+ [ Solution: Guidance for Optimizing Deep Learning Workloads for Sustainability on AWS](https://aws.amazon.com/solutions/guidance/optimizing-deep-learning-workloads-for-sustainability-on-aws/)
+ [Well-Architected Lab - Rightsizing Recommendations](https://wellarchitectedlabs.com/cost/100_labs/100_aws_resource_optimization/)
+ [Well-Architected Lab - Rightsizing with Compute Optimizer](https://wellarchitectedlabs.com/cost/200_labs/200_aws_resource_optimization/)
+ [Well-Architected Lab - Optimize Hardware Patterns and Observe Sustainability KPIs](https://wellarchitectedlabs.com/sustainability/200_labs/200_optimize_hardware_patterns_observe_sustainability_kpis/)
+ [ Well-Architected Lab - Migrating Services to Graviton ](https://www.wellarchitectedlabs.com/sustainability/100_labs/100_migrate_services_to_graviton/)
# SUS05-BP03 Use managed services
Use managed services to operate more efficiently in the cloud.
**Common anti-patterns:**
+ You use Amazon EC2 instances with low utilization to run your applications.
+ Your in-house team only manages the workload, without time to focus on innovation or simplifications.
+ You deploy and maintain technologies for tasks that can run more efficiently on managed services.
**Benefits of establishing this best practice:**
+ Using managed services shifts the responsibility to AWS, which has insights across millions of customers that can help drive new innovations and efficiencies.
+ Managed service distributes the environmental impact of the service across many users because of the multi-tenet control planes.
**Level of risk exposed if this best practice is not established:** Medium
## Implementation guidance
Managed services shift responsibility to AWS for maintaining high utilization and sustainability optimization of the deployed hardware. Managed services also remove the operational and administrative burden of maintaining a service, which allows your team to have more time and focus on innovation.
Review your workload to identify the components that can be replaced by AWS managed services. For example, [Amazon RDS](https://aws.amazon.com/rds/), [Amazon Redshift](https://aws.amazon.com/redshift/), and [Amazon ElastiCache](https://aws.amazon.com/elasticache/) provide a managed database service. [Amazon Athena](https://aws.amazon.com/athena/), [Amazon EMR](https://aws.amazon.com/emr/), and [Amazon OpenSearch Service](https://aws.amazon.com/opensearch-service/) provide a managed analytics service.
**Implementation steps**
1. Inventory your workload for services and components.
1. Assess and identify components that can be replaced by managed services. Here are some examples of when you might consider using a managed service:
[\[See the AWS documentation website for more details\]](http://docs.aws.amazon.com/wellarchitected/2023-10-03/framework/sus_sus_hardware_a4.html)
1. Identify dependencies and create a migrations plan. Update runbooks and playbooks accordingly.
+ The [AWS Application Discovery Service](https://aws.amazon.com/application-discovery/) automatically collects and presents detailed information about application dependencies and utilization to help you make more informed decisions as you plan your migration
1. Test the service before migrating to the managed service.
1. Use the migration plan to replace self-hosted services with managed service.
1. Continually monitor the service after the migration is complete to make adjustments as required and optimize the service.
## Resources
**Related documents:**
+ [AWS Cloud Products ](https://aws.amazon.com/products/)
+ [AWS Total Cost of Ownership (TCO) Calculator ](https://calculator.aws/#/)
+ [Amazon DocumentDB](https://aws.amazon.com/documentdb/)
+ [Amazon Elastic Kubernetes Service (EKS)](https://aws.amazon.com/eks/)
+ [Amazon Managed Streaming for Apache Kafka (Amazon MSK)](https://aws.amazon.com/msk/)
**Related videos:**
+ [ Cloud operations at scale with AWS Managed Services](https://www.youtube.com/watch?v=OCK8GCImWZw)
# SUS05-BP04 Optimize your use of hardware-based compute accelerators
Optimize your use of accelerated computing instances to reduce the physical infrastructure demands of your workload.
**Common anti-patterns:**
+ You are not monitoring GPU usage.
+ You are using a general-purpose instance for workload while a purpose-built instance can deliver higher performance, lower cost, and better performance per watt.
+ You are using hardware-based compute accelerators for tasks where they’re more efficient using CPU-based alternatives.
**Benefits of establishing this best practice:** By optimizing the use of hardware-based accelerators, you can reduce the physical-infrastructure demands of your workload.
**Level of risk exposed if this best practice is not established:** Medium
## Implementation guidance
If you require high processing capability, you can benefit from using accelerated computing instances, which provide access to hardware-based compute accelerators such as graphics processing units (GPUs) and field programmable gate arrays (FPGAs). These hardware accelerators perform certain functions like graphic processing or data pattern matching more efficiently than CPU-based alternatives. Many accelerated workloads, such as rendering, transcoding, and machine learning, are highly variable in terms of resource usage. Only run this hardware for the time needed, and decommission them with automation when not required to minimize resources consumed.
## Implementation steps
+ Identify which [accelerated computing instances](https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/accelerated-computing-instances.html) can address your requirements.
+ For machine learning workloads, take advantage of purpose-built hardware that is specific to your workload, such as [AWS Trainium](https://aws.amazon.com/machine-learning/trainium/), [AWS Inferentia](https://aws.amazon.com/machine-learning/inferentia/), and [Amazon EC2 DL1](https://aws.amazon.com/ec2/instance-types/dl1/). AWS Inferentia instances such as Inf2 instances offer up to [50% better performance per watt over comparable Amazon EC2 instances](https://aws.amazon.com/machine-learning/inferentia/).
+ Collect usage metric for your accelerated computing instances. For example, you can use CloudWatch agent to collect metrics such as `utilization_gpu` and `utilization_memory` for your GPUs as shown in [Collect NVIDIA GPU metrics with Amazon CloudWatch](https://docs.aws.amazon.com/AmazonCloudWatch/latest/monitoring/CloudWatch-Agent-NVIDIA-GPU.html).
+ Optimize the code, network operation, and settings of hardware accelerators to make sure that underlying hardware is fully utilized.
+ [Optimize GPU settings](https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/optimize_gpu.html)
+ [GPU Monitoring and Optimization in the Deep Learning AMI](https://docs.aws.amazon.com/dlami/latest/devguide/tutorial-gpu.html)
+ [Optimizing I/O for GPU performance tuning of deep learning training in Amazon SageMaker AI](https://aws.amazon.com/blogs/machine-learning/optimizing-i-o-for-gpu-performance-tuning-of-deep-learning-training-in-amazon-sagemaker/)
+ Use the latest high performant libraries and GPU drivers.
+ Use automation to release GPU instances when not in use.
## Resources
**Related documents:**
+ [Accelerated Computing](https://aws.amazon.com/ec2/instance-types/#Accelerated_Computing)
+ [ Let's Architect\$1 Architecting with custom chips and accelerators ](https://aws.amazon.com/blogs/architecture/lets-architect-custom-chips-and-accelerators/)
+ [ How do I choose the appropriate Amazon EC2 instance type for my workload? ](https://aws.amazon.com/premiumsupport/knowledge-center/ec2-instance-choose-type-for-workload/)
+ [Amazon EC2 VT1 Instances](https://aws.amazon.com/ec2/instance-types/vt1/)
+ [ Choose the best AI accelerator and model compilation for computer vision inference with Amazon SageMaker AI ](https://aws.amazon.com/blogs/machine-learning/choose-the-best-ai-accelerator-and-model-compilation-for-computer-vision-inference-with-amazon-sagemaker/)
**Related videos:**
+ [ How to select Amazon EC2 GPU instances for deep learning ](https://www.youtube.com/watch?v=4bVrIbgGWEA)
+ [Deploying Cost-Effective Deep Learning Inference](https://www.youtube.com/watch?v=WiCougIDRsw)