The Six Pillars, One by One
~16 min read
A quick-reference walkthrough of what each of the six pillars actually asks.
Operational Excellence — Can you run and monitor your system effectively, and continuously improve your operational processes? Key ideas: infrastructure as code (CloudFormation/CDK) instead of manual changes, small frequent reversible deployments over large risky ones, and learning from operational failures via blameless post-incident reviews.
Security — Is data and infrastructure protected, with access controlled and tracked? Key ideas: least-privilege IAM, encryption in transit and at rest (KMS/ACM), network isolation (VPC, security groups, private subnets), and detective controls (CloudTrail, GuardDuty) to notice when something's wrong.
Reliability — Does the system recover from failure and meet demand? Key ideas: multi-AZ (and where justified, multi-Region) design, automated recovery (Auto Scaling replacing unhealthy instances), and actually testing failure scenarios rather than assuming redundancy works until it's needed.
Performance Efficiency — Are you using the right resources efficiently for the workload's actual needs? Key ideas: choosing the right instance/database type rather than defaulting to the biggest, using caching (CloudFront/ElastiCache) to reduce load on origin systems, and monitoring real usage to keep adjusting rather than a one-time choice.
Cost Optimization — Are you avoiding unnecessary spend while meeting requirements? Key ideas: right-sizing instances, Reserved/Savings Plans for steady-state workloads, S3 lifecycle policies to tier cold data to cheaper storage, and tagging resources so cost is attributable and visible.
Sustainability — Are you minimizing the environmental impact of your workload? Key ideas: improving utilization efficiency (unused capacity wastes energy as well as money), choosing managed services (AWS optimizes hardware utilization across many customers more efficiently than most individual companies could), and considering Region choice's energy mix where it doesn't conflict with other requirements.
💬 Deep Dive with AI
Key points
- •Operational Excellence: infrastructure as code, small reversible changes, learn from incidents
- •Security: least privilege, encryption, network isolation, detective controls
- •Reliability: multi-AZ/Region, automated recovery, actually test failure scenarios
- •Performance Efficiency: right resource for the job, caching, monitor and iterate
- •Cost Optimization: right-sizing, Reserved/Savings Plans, lifecycle policies, tagging
- •Sustainability: utilization efficiency, managed services, Region energy mix