LOCAL PLANNING TOOL

MQTT device fleet capacity and cost calculator

Turn fleet size, publish frequency, payload size, fan-out, online share, and reconnect batches into a transparent monthly capacity estimate.

Fleet assumptions

Inputs stay in this browser tab. They are not uploaded, saved, or added to the URL.

All device records in the fleet, whether online or offline.

Average publishes from each online device. Use 0.1 for one message every 10 minutes.

Application payload size before MQTT, TLS, and network overhead.

Use 0 when publications have no matching subscriber and 2 when two copies are delivered on average.

Expected share connected and publishing at the same time.

Devices expected to reconnect in one recovery wave; this cannot exceed total devices.

Capacity estimate

Review the example assumptions, then calculate to see connections, publications, deliveries, message units, and reconnect guidance.

Transparent arithmetic

Formulas, units, and rounding

Every result uses the formulas shown here. The 30-day values are a readable baseline; the message-unit range covers 28–31 days.

  1. 01

    Concurrent connections

    ceil(devices × online %)

    Fractional connections round up because a connected client is a whole connection.

  2. 02

    Published messages

    connections × publishes/minute × 1,440 × calendar days

    The rate applies only to the estimated online connections. One day contains 1,440 minutes.

  3. 03

    Subscriber deliveries

    publishes × average subscriber fan-out

    Fan-out is an average, so estimated deliveries may be fractional before display formatting.

  4. 04

    Metered message units

    (publishes + deliveries) × max(1, ceil(payload bytes ÷ 512))

    Each publish and delivery copy is metered separately. Payloads round up to 512-byte units, with at least one unit for an empty payload.

Local and privacy-bounded

Calculation happens in browser memory. Inputs are not sent to an API, written to storage, or encoded in a share URL. Analytics receives only predefined connection, message, and reconnect bands.

Estimate, then validate

The result does not model MQTT packet overhead, TLS handshakes, QoS retries, retained-message replays, session queues, network loss, processing latency, or regional failure recovery. It cannot replace a representative load test.

Continue from estimate to architecture

Compare managed and self-hosted operations, review scaling practices, and confirm the current public pricing and message-accounting rules.

Review current Pricing