I returned to AWS and was reminded why I left

Original Article ↗ Hacker News Discussion ↗

Many engineers say returning to Amazon Web Services now feels like re‑entering an overcomplicated, adversarial ecosystem: pricing is opaque, billing “footguns” are easy to hit, and core tasks like email, IAM and support can be brittle or slow, especially for smaller users. Commenters contrast AWS’s early innovation and still-strong core (EC2, S3, RDS, IAM) with a sprawl of confusing services, lock‑in incentives, and serverless patterns that can become unexpectedly expensive at scale. A recurring theme is that for many startups and personal projects, simpler VPS providers, platforms like Vercel, Cloudflare, or DigitalOcean, or even basic self‑hosting often deliver lower cost and complexity, while AWS remains primarily attractive for large enterprises with deep expertise and budgets.

Cloud vs VPS / Self‑Hosting

  • Many argue typical startups and small apps can run on a single VPS or a few dedicated servers (Hetzner, DigitalOcean, etc.) for a fraction of hyperscaler cost.
  • Several anecdotes of companies cutting cloud bills from tens of thousands to a few hundred dollars by moving off managed services to simpler VMs and OSS (Postgres, Redis, Prometheus, Grafana, etc.).
  • Counterpoint: critics say this often trades away redundancy, durability, and managed logging/monitoring; risk profile changes, not eliminated.

AWS Complexity & IAM

  • Recurrent theme: AWS feels overengineered for hobby projects; console and IAM are intimidating and “enterprise first.”
  • Others respond that you can still just run a basic EC2 VM, and that IAM, STS, and core services are actually well-designed compared to other clouds—just complex by nature.
  • Some see the console UX and lack of inline pricing as “adversarial” design.

Billing, Pricing, and Footguns

  • Strong frustration with opaque pricing, egress fees, and lack of hard spending limits.
  • AWS credits and free tiers encourage heavy use of proprietary services; when credits expire, bills can spike dramatically.
  • Some view complex billing as intentional revenue optimization; others say it’s inherent to usage-based models.

Serverless & Managed Services

  • Split opinions on Lambda, DynamoDB, API Gateway, ElastiCache:
    • Fans praise Lambda and DynamoDB as cheap, low‑ops, especially at low or spiky volumes.
    • Critics highlight cold starts, configuration complexity, “nickel‑and‑diming” per request/unit, and poorer performance vs self‑hosted Redis/DBs.
  • Consensus: great for certain patterns, dangerous if misused or treated like generic SQL/VMs.

Open Source & AWS Forks

  • Debate over AWS’s forking and monetizing of popular OSS (Elasticsearch/OpenSearch, Redis/Valkey, etc.).
  • One camp blames AWS for “eating the ecosystem’s lunch,” forcing projects into restrictive licenses.
  • Another camp says these projects chose permissive licenses and later changed business models; competition on hosting is seen as legitimate.

Alternatives & Ecosystem

  • Strong nostalgia for Heroku-style simplicity; praise for modern PaaS and edge platforms (Vercel, Render, Fly, Cloudflare, DigitalOcean).
  • Mixed views on GCP and Azure: some find GCP IAM and VMs simpler; others highlight random bans and quota issues, or Azure security incidents.
  • Several note that AWS persists largely due to enterprise procurement friction, existing expertise, and generous startup credits, not inherent simplicity.