The Modern AWS Data Stack in 2026: Services, Patterns, and Costs
S3 as the foundation, Glue for ETL, EMR for Spark, Redshift for warehousing — with real cost benchmarks.
Get one deep-dive article every Saturday. No spam. Unsubscribe anytime.
AWS has the most mature set of managed data services of any cloud provider. The challenge isn't finding a service — it's choosing the right combination and understanding how costs compound. This is a practical guide to what the modern AWS data stack looks like, what it costs, and when to choose each service.
The Foundation: S3
Every AWS data platform starts with S3. It's the cheapest, most durable object storage on the planet — 11 nines of durability, $0.023/GB/month.
Key design decisions:
Processing: Glue vs EMR
Use Glue when: Simple ETL, no cluster management, infrequent jobs.
Use EMR when: Fine-grained Spark config, long-running jobs, cost matters (EMR Spot is 70% cheaper than Glue).
EMR Serverless is the middle ground: Spark on demand with no cluster management.
Get unlimited access to all Cloud tutorials & animated diagrams
1,000+ animated tutorials · 100+ AI tools · Mock interviews · Certifications
Real Cost Benchmark: 10TB/day Pipeline
| Service | Monthly Cost |
|---|---|
| S3 Storage (90 days, 900TB) | ~$20,000 |
| Kinesis Firehose | ~$900 |
| EMR on Spot (8 r5.4xlarge × 6h/day) | ~$1,200 |
| Athena (5TB scanned/day) | ~$750 |
| Redshift RA3 (ra3.4xlarge × 2) | ~$2,500 |
| **Total** | **~$25,350/month** |
With S3 Intelligent-Tiering, Spot, and partition pruning, this drops to ~$14,000/month.
Cost Optimization Priorities
Get unlimited access to all Cloud tutorials & animated diagrams
1,000+ animated tutorials · 100+ AI tools · Mock interviews · Certifications