Should I size for average traffic?

Use expected peak or protected peak values for safer planning.

Is one redundancy instance always enough?

Not always. Apply availability and failure-domain requirements.

How to find per-instance capacity?

From controlled load tests with realistic endpoints and payloads.

Infrastructure sizing calculator

Estimate minimum and recommended instance counts based on expected load, per-instance throughput capacity and safety margin.

Calculator

Expected RPSCapacity per instance (RPS)Safety margin (%)Add one redundancy instance

Sizing output

Minimum instances

Recommended instances

Total provisioned capacity

1,750 RPS

Estimated headroom

550 RPS

Recommendation includes selected safety margin and optional redundancy instance.

What this calculator solves

It converts traffic targets into practical instance counts with room for variability and failover.

Recommended safety margin

instances_min = ceil(expected_rps / capacity_per_instance)
instances_margin = ceil(expected_rps * (1 + margin_pct/100) / capacity_per_instance)
instances_recommended = instances_margin + redundancy

Use margin to absorb variability and include redundancy if one instance failure must not breach throughput targets.

Example

For 1200 RPS target with 250 RPS per instance and 20% margin, base recommendation is 6 instances. With redundancy enabled, recommendation becomes 7.

FAQ

Should I size with average or p95 load?

Use expected peak load for operational safety.

Can one instance have variable capacity?

Yes. This calculator assumes a stable tested average.

Is one redundancy instance enough?

It depends on your availability targets and fault domains.