System Deployment Model
MAPS is deployed as one enterprise hub connected to many edge sites (1 → 10,000).
Each edge site runs MAPS alongside an IoT modem; the enterprise runs MAPS connected to Viasat/Inmarsat/Orbcomm REST APIs.
Roles at a Glance
| Area | What it runs | Key responsibilities |
|---|---|---|
| Edge (Remote sites) | MAPS Edge + IoT Modem (OGx/IDP) | Buffering, filtering/selectors/ML, packaging, chunking, retry/back-off, GNSS/visibility awareness |
| Satellite Transport | L-band GEO + ground segment | Store-and-forward, queued delivery, variable RTT |
| Enterprise (Datacenter/Cloud) | MAPS Enterprise | REST/API polling, dedupe/ack tracking, namespace routing to MQTT/AMQP/NATS/REST, fleet scale ops |
End-to-End Topology (1 → 10,000)
Data Flow (Concise)
- Ingest at Edge: Sensors → MAPS Edge topics (schema-aware).
- Reduce: Selectors/ML filter events; package → CRC → compress → encrypt → chunk.
- Uplink: Sent when satellite visible; retries with back-off.
- Transit: Satellite + ground segment store-and-forward.
- Enterprise Intake: MAPS Enterprise polls REST, reassembles, validates, republishes to enterprise namespaces.
- Consume: Apps use MQTT/AMQP/NATS/REST without satellite quirks.
Scale & Tuning
- Edge fan-out: 1 → 10,000 sites.
- Polling (enterprise): Incoming ≥10 s, Outgoing ≥15 s (tune per fleet size/API cost).
- Packaging: Prefer last 1 or last n per topic; dedupe aggressively.
- Latency budget: Seconds (not ms). Size timeouts accordingly.
- Back-pressure: Cap inflight per device; stagger flush windows.
Operational Notes
- Resilience: Edge buffers during outages; enterprise dedupes across poll cycles.
- Namespace discipline: Keep predictable mappings for routing/ACLs and per-device topics.
- Security: Shared secrets and payload encryption end-to-end; rotate keys on a schedule.
- Observability:
- Edge: visibility state, GNSS, queue depth, retry counts, chunk loss.
- Enterprise: poll lag, inflight per device, dedupe hits, reassembly failures.
- Updates: Push selector/namespace/ML updates from enterprise; no device firmware changes required for most policy tweaks.
When This Model Fits
- Maritime, mining, utilities, transport—anywhere with intermittent or costly links.
- Fleets from hundreds to tens of thousands of devices.
- Need for protocol-agnostic apps (MQTT/AMQP/NATS/REST) without satellite specifics leaking in.
Implementation Pointers
- Edge configuration: Set conservative send windows, compression cutoff, max buffer size; enable selectors/ML where it reduces bytes.
- Enterprise configuration: Set poll intervals by fleet size; cap inflight per device; enforce minimums.
- Testing: Simulate loss/dup/latency; validate chunk reassembly and CRC; confirm namespace mappings end-to-end.