From capacity to dollars
Cost modeling maps your resource estimates onto cloud prices to predict the bill. It connects engineering choices to money, which is what gets a design approved.
The main line items
- Compute: instances per hour, scaled by server count and uptime.
- Storage: gigabytes per month, multiplied by replication.
- Egress: bytes leaving the cloud, often the surprise that dominates media systems.
- Managed services: databases and queues priced per throughput or per request.
Levers that cut cost
- Reserved or committed pricing trades flexibility for a discount.
- Autoscaling avoids paying peak rates around the clock.
- Caching and CDNs cut egress, frequently the largest variable cost.
Egress catches many teams off guard because it scales with traffic and is billed per gigabyte, unlike the inbound direction which is usually free.
Key idea
Cost models sum compute, storage, egress, and managed services, with egress and committed pricing the biggest levers.