Workload Catalog
Find your workload. See it proven.
Aggregation and reduction workloads that gate on throughput at correctness — hitting your rate while the result stays exactly correct no matter the arrival order or how many cores feed it. Each archetype below is checked against the silicon-measured parallel-bank engine. Pick the one that's yours.
Representative archetypes — not named customers. Verdicts use measured engine facts (AX7020, 100 MHz, byte-identical across 1/2/4/8 banks).
Recognize yours? Become a founding design partner → — we run it live on silicon behind a conditional LOI you only honor if we hit your bar.
High-cardinality metric rollup
Counters, gauges and histogram buckets from thousands of agents, rolled up exactly per series.
25 M/s required → 800 M/s measured, byte-identical · see the proof →
Real-time per-entity counters
Per-IP / per-user / per-card event counters for fraud and rate-limit decisions, exact under concurrency.
100 M/s required → 800 M/s measured, byte-identical · see the proof →
Impression & click aggregation
Per-campaign impression, click and spend counters aggregated across bid/serve nodes in real time, exact for pacing and capping.
200 M/s required → 800 M/s measured, byte-identical · see the proof →
Fleet telemetry rollup
Per-device telemetry deltas from a large fleet, rolled up exactly despite late, out-of-order delivery — no windowing, no watermarks.
40 M/s required → 800 M/s measured, byte-identical · see the proof →
Distinct-count cardinality rollup
Unique users / devices / sessions estimated with sketches and merged across shards and windows — byte-identical regardless of merge order, no serialized reducer.
20 M/s required → 800 M/s measured, byte-identical · see the proof →
Replica state convergence
Replicas converge to identical state by exchanging deltas — no coordinator, no consensus, a fraction of the bandwidth.
160 B vs 2048 B/tick · converges with no coordinator · see the proof →
Offline-first client sync
Field / mobile clients work offline, then converge on reconnect by exchanging deltas — no server merge pass, no last-writer-wins data loss, a fraction of the bandwidth.
160 B vs 2048 B/tick · converges with no coordinator · see the proof →