Parallelized query execution across nodes, cores, and SIMD lanes

Parallelization levelDrag to exploreCompute nodes

Compute nodesCPU coresSIMD lanes

Data is processed in parallel by multiple compute nodes.

Data is processed in parallel by multiple CPU cores.

Data is processed in parallel by SIMD lanes.

The numbers 1–48 identify illustrative data elements, not benchmark values. Node 2 receives elements 17–32. Its core 2 receives elements 21–24, which are assigned to four SIMD lanes for one instruction. Each lane applies the same multiplication by 10: 21, 22, 23 and 24 become 210, 220, 230 and 240. Multi-node processing refers to distributed queries; SIMD applies to supported operations.