Capacity planning at planet scale
Forecast when growth hits saturation, choose headroom for queueing, failures and deploys, find the first component to break, and read a Scale-mode breaking-point sweep.
30 min6 steps
By the end you can
- Forecast when compounding growth reaches a system's saturation point
- Choose a utilisation target that leaves room for queueing delay, failures and deploys
- Find which component saturates first as load grows, and at what multiple
- Interpret a breaking-point sweep and choose between independent levels and a continuous ramp
Plan for the peak you will have
Average
40k/s
Daily peak
120k/s
Launch (+20% at peak)
144k/s
Compounding growth arrives sooner
- Today1 x
- Month 61.6 x
- Month 122.5 x
- Month 246.3 x
Loading question…
Why you never plan for 100%
- 50% busy2 x service time
- 80% busy5 x service time
- 90% busy10 x service time
The first thing to break
Loading question…
Sweep the load in Scale mode
Independent levels
- Fresh after warm-up
- What a warmed system sustains
Continuous ramp
- State carries over
- Shows retry storms, autoscale lag, cache churn
Check your mastery, then find the breaking point
Sweep
1B requests/day
independent levels
Then
continuous ramp
compare ladders
Score 80% or more. Open the no-cache design, use the Scale tab, note the first rung, try its remedy, rerun as a ramp.
Mastery check
4 questions on fresh variants. One attempt each, no time limit. Score 80% or more to master the lesson.
Answer this step to continue.