IoT / Edge
Fleet telemetry rollup
Per-device telemetry deltas from a large fleet, rolled up exactly despite late, out-of-order delivery — no windowing, no watermarks.
Measured 2026-06-21 on ALINX/HamGeek AX7020 (XC7Z020-2CLG484); reproduce with abench 65537 0xDEADBEEF0BADF00D. Throughput is the engine's accumulation rate (the aggregation step); end-to-end depends on your ingest path, which ATOMiK's order-independence frees from global ordering / locks.
This is you if
- You run a large connected-device fleet (cellular / LoRa / NB-IoT).
- You roll up per-device / per-metric usage counters for billing, SLA, or alerting.
- Late and out-of-order arrivals force buffering, watermarks, or reprocessing.
The workload
- Millions of devices emit per-metric deltas: usage counters, event tallies, totals.
- Transport is lossy and high-latency; deltas arrive wildly out of order and delayed.
- Rollups must be exact per device / metric per window — billing and SLA depend on it.
- Devices are concurrent and uncoordinated.
Today — serialized behind a lock
Exact rollups under late, out-of-order arrival usually means per-device windowing with watermarks plus a serialized commit path — adding latency, memory, and reprocessing whenever late data shows up after the window closed.
With ATOMiK — order-independent, lock-free
Devices feed deltas into a shared accumulator in any order, at any time. The rollup is order-independent and byte-identical, so a delta that arrives an hour late simply accumulates into the exact result — no sequencing, no watermark stalls, no reprocessing.
- Late and out-of-order data accumulates correctly — no watermarks, no reprocessing.
- No per-device sequencing / windowing lock on the commit path.
- Exact, byte-identical rollups for billing and SLA.
- Linear scaling across the fleet; devices never coordinate.