Agent Skills
› cosmicstack-labs/mercury-agent-skills
› cloud-architecture
cloud-architecture
GitHub提供多云架构设计指南，涵盖VPC多租户设计、高可用部署及灾难恢复策略。结合可靠性、安全、成本与性能五大支柱，给出最佳实践，助力构建弹性、经济且易运维的云系统。
Trigger Scenarios
设计高可用云架构
规划VPC网络拓扑
制定灾难恢复方案
优化云资源成本
Install
npx skills add cosmicstack-labs/mercury-agent-skills --skill cloud-architecture -g -y
SKILL.md
Frontmatter
{
"name": "cloud-architecture",
"metadata": {
"tags": [
"cloud",
"architecture",
"aws",
"gcp",
"azure",
"ha",
"dr"
],
"author": "cosmicstack-labs",
"version": "1.0.0",
"category": "devops"
},
"description": "Multi-cloud, VPC design, high availability, disaster recovery, and cost optimization"
}
Cloud Architecture
Design resilient, cost-effective cloud architectures.
Design Pillars
| Pillar | Focus | Pattern |
|---|---|---|
| Reliability | Fault tolerance, recovery | Multi-AZ, redundancy, auto-healing |
| Security | Defense in depth | IAM, encryption, network segmentation |
| Cost | Efficiency, waste reduction | Right-sizing, reserved instances, autoscaling |
| Performance | Speed, scalability | CDN, caching, horizontal scaling |
| Operational | Manageability, automation | IaC, CI/CD, runbooks |
VPC Design
Multi-Tier Architecture
Internet → WAF → ALB → App Tier (private) → DB Tier (private)
↕
Auto Scaling
Best Practices
- Use multiple AZs (minimum 3)
- Private subnets for apps and databases
- Public subnets only for load balancers and bastions
- NACLs for subnet-level rules, Security Groups for instance-level
- VPC peering or Transit Gateway for multi-VPC
High Availability
- Compute: Multi-AZ Auto Scaling groups, spot + on-demand mix
- Database: Multi-AZ RDS, Aurora read replicas
- Storage: S3 (11 9's durability), EBS snapshots
- DNS: Route53 with health-check based failover
- CDN: CloudFront/CloudFlare for global distribution
Disaster Recovery
| Strategy | RTO | RPO | Cost |
|---|---|---|---|
| Backup & Restore | Hours | 24h | $ |
| Pilot Light | Minutes | 1h | $$ |
| Warm Standby | Seconds | Minutes | $$$ |
| Active-Active | Seconds | Seconds | $$$$ |
DR Plan
- Document runbooks for each scenario
- Test failover quarterly
- Automate recovery with IaC
- Store backups in separate region
- Encrypt everything
Cost Optimization
- Right-size instances (use Compute Optimizer)
- Reserved Instances / Savings Plans for steady state
- Spot instances for fault-tolerant workloads
- S3 lifecycle policies (transition to Glacier)
- Delete unused resources (EBS, EIP, ELB)
- Monitor with Cost Explorer + budgets + alerts
Version History
- 38e2523 Current 2026-07-05 19:38
