20.11 Capacity, Cost & FinOps
Cost optimization in Datasphere is an architecture problem, not simply a finance exercise.
Main cost drivers
Depending on contract and service configuration, evaluate:
- storage consumption
- compute workload
- data movement
- integration patterns
- refresh frequency
- persistence choices
- concurrency
Capacity model
Track:
Baseline + Growth + Peak + Safety Margin
Do not size only for today's average load.
Architecture levers
Reduce unnecessary data movement
Prefer federation when latency, source load and governance permit it.
Persist selectively
Persist when it materially improves workload behavior or isolates expensive repeated computation.
Control refresh frequency
Align frequency with business freshness requirements rather than maximum possible frequency.
Remove unused assets
Retire stale flows, tables, views and schedules through governed lifecycle processes.
FinOps dashboard
Track:
- capacity trend
- workload trend
- top consumers
- unused objects
- refresh frequency
- cost per business capability where measurable
Architect trade-off
The cheapest design is not necessarily the best design.
Optimize:
Business value / total operating cost
while respecting reliability, security and performance.