Project · Multi-Cloud Cost Comparison

project

hard

module project

Ship something real. Submit your work when you're done.

Brief

Price the same 3-tier architecture (web + app + database) on AWS, Azure, GCP, and DigitalOcean using each provider's official calculator. Document every assumption, normalize the SKUs across clouds, and produce a reproducible markdown cost table that another builder could rebuild from your notes alone.

Deliverables

An `architecture.md` describing the reference 3-tier stack: 2x small web instances behind a load balancer, 1x medium app instance, 1x small managed Postgres (single-AZ), 100GB object storage, 1TB egress/month. Region: closest to your location on each cloud.
A `costs.md` markdown table with one row per provider and columns for compute, database, storage, egress, load balancer, and monthly total in both USD and INR.
A `assumptions.md` listing every assumption you made when SKUs did not map cleanly (e.g. AWS t3.medium ~ Azure B2s ~ GCP e2-medium ~ DO basic-2vcpu-4gb), with a one-line justification each.
A short `reflection.md` (200–300 words) answering: which provider was cheapest, which was most expensive, what surprised you, and which provider you would pick for this workload and why.
Calculator share-links (or screenshots) for all four providers as evidence — paste them in a `calculator-links.md`.

How we grade it

All four providers priced for the same logical architecture (compute count, storage size, egress, DB tier all match within ±10%).
Cost table is sortable and shows monthly total in both USD and INR — INR conversion uses a single documented exchange rate, not different rates per provider.
Every assumption that affects price by more than 10% is called out explicitly in `assumptions.md` (instance family, reserved vs on-demand, egress tier, storage class).
The reflection identifies at least one non-price factor (region availability, free tier, support cost, lock-in) that would change your final pick.
Calculator share-links are valid and resolve to the exact configuration you priced — a reviewer can click them and reproduce your number.

Hints

AWS — use https://calculator.aws. Pick "Add service", then EC2, RDS, S3, ELB, Data Transfer. Save as a "share link" at the top right.

Azure — use https://azure.microsoft.com/en-us/pricing/calculator/. Add Virtual Machines, Azure Database for PostgreSQL Flexible Server, Storage Accounts, Load Balancer, and Bandwidth. Export to Excel or save with a Microsoft account.

GCP — use https://cloud.google.com/products/calculator. Compute Engine, Cloud SQL for PostgreSQL, Cloud Storage, Cloud Load Balancing, Network egress. Click "Share" to get a permalink.

DigitalOcean — use https://www.digitalocean.com/pricing/calculator. Droplets, Managed Databases, Spaces, Load Balancer, Bandwidth. Flat pricing means fewer knobs — that's the point.

Normalization tip: pick a vCPU + RAM target (e.g. "2 vCPU, 4 GB, ~$30-40/mo") and find the closest SKU on each cloud rather than chasing exact equivalence.

Expected output

| Provider | Compute | Database | Storage | Egress | LB | Total (USD) | Total (INR) | |--------------|---------|----------|---------|--------|-------|-------------|-------------| | AWS | $62.40 | $51.10 | $2.30 | $90.00 | $18.40| $224.20 | INR 18,712 | | Azure | $58.40 | $48.50 | $2.10 | $86.70 | $18.25| $213.95 | INR 17,856 | | GCP | $54.10 | $46.80 | $2.00 | $80.00 | $18.00| $200.90 | INR 16,775 | | DigitalOcean | $48.00 | $30.00 | $5.00 | included| $12.00| $95.00 | INR 7,929 | *Exchange rate used: 1 USD = 83.46 INR (RBI reference, 2025-04-30)* *Region: us-east (AWS/Azure/GCP us-east-1/eastus/us-east1; DO NYC3)* *Workload: 2x 2vCPU/4GB web, 1x 4vCPU/8GB app, 1x 2vCPU/4GB Postgres single-AZ, 100GB object storage, 1TB egress*

Stretch goals

Re-price the same architecture under 1-year reserved / committed-use on AWS, Azure, and GCP and add a "Reserved (1yr)" column — note the breakeven utilization.

Add a fifth provider (Hetzner, Linode/Akamai, or OVH) to the table and see how the hyperscalers compare to a budget/EU provider.

Model 10x scale (20 web instances, 10 app instances, 10TB storage, 100TB egress) and observe which provider's pricing curve is most punishing — egress is usually the answer.

Build a small Python script that calls each provider's pricing API directly and regenerates the table without you re-clicking the calculator.

Add a "Total Cost of Ownership" column that includes one engineer-hour per month for ops on each provider — make a defensible estimate of whether the cheapest cloud is actually cheapest end to end.