Bulk onboarding with Agent Smith

When a new agent registers with Rewst, the workflow hits Azure IoT Hub’s identity registry up to five times.

Two queries → create → get → tag‑update

On a standard‑1 (S1) hub, that registry is capped at 100 identity operations per minute, per unit.

100 ops ÷ 5 ops / agent ≈ **20 agents / min / unit**

Important: Plan your own throttle or scale‑up Rewst deliberately leaves rate‑limiting in your hands so you can tune roll‑outs to match your Azure budget. A large, unthrottled burst will exceed IoT Hub limits, pile up HTTP 429 retries, and slow down all of your Rewst workflows during onboarding. To avoid this, either add jitter/batching in your RMM script, or do a temporary IoT Hub scale‑up before the push.

What is an IoT hub unit?

A unit is simply a slice of IoT Hub capacity you can dial up or down. All per‑minute limits scale linearly with both units and SKU size.

SKU	Identity ops / min / unit	Agents / min / unit, 5‑call workflow
S1 / S2	100	~20
S3	5 000	~1 000

---

Scenarios

Choose one	How to do it	Throughput	Cost example*
Scale the hub (fast)	• bash az iot hub update -n <hub> --sku S3 --unit 2 • Run 5 min before deploy; scale back to S1 --unit 1 when done.	~2 000 agents/min	S3 ≈ $3.47 per unit‑hour → 2 units × 4 h ≈ $28
Throttle in RMM (no Azure cost)	• Deploy ≤ 50 agents / min / S1‑unit. • Add jitter: Start‑Sleep (Get‑Random 5 3600) before install.	Current hub limit (~20 agents/min)	Longer total time; more scripting effort

*US‑East pricing, rounded

Rules to follow

• Default S1 × 1 → ≈ 20 agents / min.

• Formula: Agents / min = (100 or 5 000) × units ÷ 5.

• Scaling is live. Existing connections stay up; a few messages may arrive out of order, which is harmless.

Pick the path that matches your timeline and budget to avoid 429s during mass onboarding.