aws-performance

AWS Performance & Cost Checklist

CloudWatch & Monitoring

• Free tier metrics enabled for all services
• Custom metrics for application-level KPIs
• Detailed monitoring enabled for production EC2/RDS
• Alarms configured with SNS → Slack notifications
• Scale down on INSUFFICIENT_DATA and ALARM states
• Log retention policies set (don't keep logs forever)
• Dashboard for key metrics (CPU, memory, latency, error rates)

RDS / Database

• Enhanced Monitoring enabled (OS-level metrics)
• Performance Insights enabled (query-level analysis)
• Connection pooling configured (max connections vs instance size)
• SSL enforced for all connections
• Read replicas for read-heavy workloads
• Automated backups with tested restore procedures
• Parameter group tuned (work_mem, shared_buffers, max_connections)
• Slow query logging enabled (> 1s threshold)
• Reserved instances for predictable workloads (40-60% savings)
• Right-sized instances — monitor CPU/memory utilization

EC2 / Compute

• No static/elastic IPs where not needed
• Auto-scaling groups with proper scaling policies
• Single scaling trigger per dimension (CPU or request count, not both)
• Terminate protection on critical instances
• Reserved instances or Savings Plans for baseline capacity
• Spot instances for fault-tolerant batch workloads
• Right-sized instances — aim for 60-80% average utilization

S3 / Storage

• Lifecycle policies for old data (→ Glacier after 90 days)
• Intelligent-Tiering for unpredictable access patterns
• CloudFront CDN for frequently accessed objects
• KMS encryption with customer-managed keys
• Versioning enabled on critical buckets
• Request metrics enabled for high-traffic buckets

ELB / Networking

• Terminate SSL on load balancer (offload from app servers)
• Pre-warm for expected traffic spikes (contact AWS support)
• Health checks configured (HTTP path, not TCP)
• Cross-zone load balancing enabled
• Connection draining enabled for graceful deploys

Caching Layers

• ElastiCache for session state and hot data
• Use configuration endpoints (not individual node addresses)
• CloudFront for static assets (CSS, JS, images)
• HTTP cache headers on API responses where appropriate
• Cache invalidation strategy documented

WAF & Security

• Rate limiting rules on all public endpoints
• IP whitelisting for admin/internal endpoints
• Managed rule groups for common attack patterns (SQLi, XSS)
• Logging enabled with adequate retention
• Environment-specific rules (stricter in production)

Cost Controls

• AWS Budgets with alerts at 50%, 80%, 100%
• Cost Explorer reviewed monthly
• Unused resources cleaned up (unattached EBS, old snapshots, idle NAT gateways)
• Right-sizing recommendations reviewed quarterly
• Tag everything — environment, service, team, cost-center
• Reserved capacity for predictable workloads

General Best Practices

• Redundancy across at least 2 AZs
• Infrastructure as Code (Terraform) — no manual console changes
• Service limits documented and increased before hitting them
• Disaster recovery tested (at least annually)
• Deployment strategy: blue-green or rolling (zero downtime)

Cost Optimization Quick Wins

Action	Typical Savings	Effort
Reserved Instances (RDS/EC2)	40-60%	Low
S3 Lifecycle policies	30-50% on storage	Low
Right-size EC2 instances	20-40%	Medium
Spot instances for batch	60-90%	Medium
CloudFront for static assets	Reduced S3 requests	Low
Remove unused EBS volumes	Direct cost reduction	Low
Intelligent-Tiering S3	Automatic savings	Low

AWS Performance Antipatterns

# BAD: Over-provisioned "just in case"
Instance: r5.4xlarge (128GB RAM)
Actual usage: 8GB average

# GOOD: Right-sized with auto-scaling
Instance: r5.large (16GB RAM) + auto-scaling policy
Scale out at 70% CPU, scale in at 30%

# BAD: Multiple scaling triggers competing
aws autoscaling put-scaling-policy --metric CPU --threshold 70
aws autoscaling put-scaling-policy --metric RequestCount --threshold 1000
# These fight each other!

# GOOD: Single primary metric per ASG
aws autoscaling put-scaling-policy --metric RequestCount --threshold 1000

Recommended External Skills

# AWS cost optimization + monitoring (5 plugins with MCP servers)
/plugin marketplace add zxkane/aws-skills
/plugin install aws-cost-ops@aws-skills

# AWS cost scanner — 163 checks across 30+ services
# From: github.com/prajapatimehul/aws-cost-scanner