Quarry tracks 9,000 trucks on one view
GPS pings, CDC from the dispatch database and a single join, served under 5 ms.
Quarry moves aggregates for construction sites across three countries. Every one of their 9,000 trucks sends a GPS ping every five seconds.
One join, always current
Pings stream in over MQTT. Orders and assignments arrive by CDC from the dispatch database. A single Basalt view joins them, so every truck carries its current job, ETA and load.
1,800 pings per second at peak.
Dispatchers query the view from a map that refreshes every second.
Drivers see the same ETA in their app.
Before Basalt our map was five minutes behind. Now it is behind by the speed of light.
Late deliveries fell 22% in the first two months, mostly because dispatchers could reroute before a truck hit traffic, not after.
More from the engine room.
How we cut p99 from 11 ms to 4.2 ms
Three changes to the read path in v4.0: declared sort keys, a per-core allocator and a planner that stopped re-sorting partitions at query time.
Exactly-once without two-phase commit
Idempotent sinks, epoch fencing and a 64-bit sequence per partition. No coordinator, no stalls.
Time travel is just a really good index
Why AS OF queries on 30 days of history cost less than one extra replica.
