One parallelized query scales across 100 nodes

1 trillion raw rows aggregated, without pre-aggregation or data reshuffling

FOR CONTEXT

One query distributed across compute nodes

Query compute89 CPU cores per node
Drag to explore

Data processing throughput

Uncompressed GB/s

Compute nodes / total CPU cores
Total throughput

Separate experiment: UK property-price aggregation · Full benchmark ↗

One trillion UK property-price rows. Parallel replicas, 89 CPU cores per node. Best measured query execution of ten runs at each node count. The chart uses proportional numeric spacing; only measured node counts can be selected.