Total Recommendations
12
Across all optimization categories
Estimated Monthly Savings
$45+
If all recommendations applied
Quick Wins
5
Low effort, immediate impact
Sources
4
Cost Optimization Hub, Compute Optimizer, Trusted Advisor, FinOps Agent
Cost Optimization Recommendations (Prioritized by Savings)
| Priority | Category | Recommendation | Resource | Est. Savings | Effort |
|---|---|---|---|---|---|
| HIGH | Commitment | Purchase Savings Plan for EC2 usage (3-year No Upfront covers $7.59/mo EC2 at ~40% discount) | Account-wide | $3.00/mo | LOW |
| HIGH | Idle Resources | Terminate unused Elastic IPs not attached to running instances | EIP | $3.65/mo | LOW |
| HIGH | Rightsizing | Downsize t2.xlarge non-production instances to t3.medium when not in active use (4 vCPU not needed 24/7) | Non-Prod EC2 | $9.50/mo | MED |
| MED | Scheduling | Stop non-production EC2 instances during nights/weekends (save 65% of runtime) | Dev/Test instances | $5.00/mo | MED |
| MED | Storage | Delete unattached EBS volumes and old snapshots (check for orphaned storage) | EBS Volumes | $2.00/mo | LOW |
| MED | S3 Lifecycle | Add S3 Lifecycle policies to transition infrequently accessed data to S3 Glacier | S3 Buckets | $1.50/mo | LOW |
| MED | Logging | Set CloudWatch Log retention policies (1-30 days for non-critical logs) | 17 Log Groups | $3.00/mo | LOW |
| LOW | Networking | Use VPC Endpoints for S3/DynamoDB to avoid data transfer charges | VPC | $0.50/mo | MED |
| LOW | Instance Type | Migrate t3.micro dashboards to Graviton t4g.micro (20% cheaper, same performance) | Dashboard EC2s | $1.50/mo | MED |
| LOW | Domain | Evaluate if all Route53 hosted zones are needed (each costs $0.50/mo) | Route53 | $0.50/mo | LOW |
| LOW | Reserved Capacity | Consider Reserved Instances for always-on t3.micro workloads (1-year commitment) | EC2 | $2.50/mo | LOW |
| LOW | Monitoring | Remove unused CloudWatch dashboards ($3/dashboard/month) | CloudWatch | $12.00/mo | LOW |
Tagging & Cost Allocation Recommendations
| Priority | Category | Recommendation | Resource | Impact | Effort |
|---|---|---|---|---|---|
| HIGH | Tagging Policy | Enforce required tags (Name, Environment, Owner, Project, CostCenter) on all resources using AWS Organizations Tag Policies | Organization-wide | Cost visibility | MED |
| HIGH | Untagged Resources | Tag all EC2 instances, EBS volumes, and RDS databases with Environment and Owner tags for accurate chargeback | EC2, EBS, RDS | Chargeback accuracy | LOW |
| HIGH | Cost Allocation | Activate cost allocation tags in the Billing Console to enable tag-based cost reports and budget filtering | Billing Console | Reporting | LOW |
| MED | Tag Automation | Deploy AWS Config rule required-tags to detect and alert on non-compliant resources at creation time | AWS Config | Prevention | MED |
| MED | SCP Enforcement | Create an SCP that denies resource creation without required tags (Environment, Owner) in production OUs | Organizations | Enforcement | MED |
| MED | Tag Inheritance | Enable tag inheritance for Auto Scaling groups so launched instances inherit group-level tags automatically | Auto Scaling | Consistency | LOW |
| LOW | Tag Hygiene | Standardize tag values (e.g., "production" vs "prod" vs "Production") using AWS Tag Editor bulk operations | All Resources | Data quality | LOW |
| LOW | Automation | Use AWS Lambda + EventBridge to auto-tag resources at creation with Creator, CreatedDate, and default Environment | Lambda | Auto-compliance | MED |
AI/ML (FinOps for AI) Recommendations
| Priority | Category | Recommendation | Resource | Impact | Effort |
|---|---|---|---|---|---|
| HIGH | SageMaker Endpoints | Delete idle SageMaker endpoints not receiving traffic. Real-time endpoints bill per hour even with zero invocations | SageMaker | $50-500/mo per endpoint | LOW |
| HIGH | Notebook Instances | Stop SageMaker notebook instances when not in use. Use lifecycle configs to auto-stop after idle timeout | SageMaker | $50-200/mo per notebook | LOW |
| HIGH | Bedrock Model Selection | Use smaller/cheaper foundation models (Haiku vs Sonnet, Titan vs Claude) for simple tasks. Match model capability to task complexity | Bedrock | 60-80% per-request savings | MED |
| MED | SageMaker Savings Plans | Purchase SageMaker Savings Plans for consistent ML workloads (1 or 3-year commitment for up to 64% discount) | SageMaker | Up to 64% savings | LOW |
| MED | Bedrock Provisioned Throughput | Use Provisioned Throughput for high-volume Bedrock inference instead of on-demand pricing to reduce per-token costs | Bedrock | 30-50% savings at scale | MED |
| MED | Training Spot Instances | Use SageMaker Managed Spot Training for fault-tolerant training jobs (up to 90% discount vs on-demand) | SageMaker | Up to 90% training cost reduction | MED |
| MED | Endpoint Auto-Scaling | Configure SageMaker endpoint auto-scaling to scale down during low-traffic periods instead of fixed instance count | SageMaker | 30-60% endpoint savings | MED |
| LOW | Bedrock Caching | Implement prompt caching for repeated Bedrock queries. Cache common responses to reduce token consumption | Bedrock | 20-40% token savings | MED |
| LOW | Multi-Model Endpoints | Consolidate multiple SageMaker models onto single multi-model endpoints to share hosting infrastructure | SageMaker | 40-70% hosting savings | HIGH |
| LOW | Serverless Inference | Move low-traffic SageMaker endpoints to Serverless Inference (scales to zero, pay only for invocations) | SageMaker | 90%+ savings for bursty workloads | MED |
AWS Cost Optimization Hub
Aggregates recommendations from Compute Optimizer, Trusted Advisor, and custom rules. Enable in Console for unified view.
AWS Compute Optimizer
ML-based rightsizing for EC2, EBS, Lambda, ECS. Requires 7+ days of metric data for accurate recommendations.
AWS Trusted Advisor
Identifies idle load balancers, underutilized EBS volumes, unassociated Elastic IPs, and low-utilization EC2 instances.
AWS FinOps Agent
AI-powered agent that investigates cost anomalies, generates optimization reports, and creates Jira tickets for action items.