Skip to content
VortexVortex

Benchmarks

Compression Speed26 chartsencode 2.7× · decode 7.4× faster than Parquet

Write Throughput

#1vortex2.74x2071 MB/s
#2parquet1.00x1044 MB/s
#3lance0.75x681 MB/s

Scan Throughput

#1vortex7.44x14.13 GB/s
#2lance1.55x3.20 GB/s
#3parquet1.00x2.91 GB/s
Geomean throughput ratio to Parquet | Aggregate throughput (higher is better)
Compression & File Size27 charts3% larger than Parquet (zstd)
TotalMBvs Arrow
#1vortex-compactTotal: 1,538.4 MB8.04x
#2parquet (zstd)Total: 1,940.6 MB6.54x
#3vortexTotal: 2,030.8 MB6.3x
#4parquet (snappy)Total: 2,710.6 MB4.62x
#5lanceTotal: 4,703.4 MB3.48x
#6arrow-ipcTotal: 12,586.1 MB1.08x
vs parquet (zstd)MinGeoMeanMax
vortex-compact: minimum 0.586x, geometric mean 0.807x, maximum 1.02x
parquet (zstd): minimum 1x, geometric mean 1x, maximum 1x
vortex: minimum 0.702x, geometric mean 1.03x, maximum 1.41x
parquet (snappy): minimum 0.999x, geometric mean 1.41x, maximum 1.59x
lance: minimum 0.868x, geometric mean 1.99x, maximum 4.26x
arrow-ipc: minimum 1.27x, geometric mean 6.06x, maximum 177x
Total file size, and geometric means of file sizes versus Arrow (higher is better) and versus parquet (zstd) (lower is better)
Random Access (NVMe)9 chartshot 75× · cold 21× faster than Parquet

Hot

#1vortex1.00x820.23 us
#2litdata2.34x1.73 ms
#3lance3.00x3.11 ms
#4arrow-ipc5.35x15.65 ms
#5parquet74.82x63.79 ms

Cold

#1vortex1.32x3.22 ms
#2lance1.89x4.68 ms
#3arrow-ipc2.17x16.20 ms
#4litdata20.18x46.53 ms
#5parquet28.13x64.56 ms
Geomean of take time ratio to fastest across every dataset (lower is better)
Random Access (S3)9 chartshot 67× · cold 2.2× faster than Parquet

Hot

#1vortex1.00x21.84 ms
#2lance25.95x164.44 ms
#3parquet67.33x394.67 ms

Cold

#1vortex1.00x251.41 ms
#2parquet2.25x554.48 ms
#3lance2.31x581.33 ms
Geomean of take time ratio to fastest across every dataset (lower is better)
Clickbench43 chartsDuckDB 1.9× · DataFusion 1.6× faster than Parquet

Performance Summary

#1duckdb:vortex1.10x3.58 s
#2datafusion:vortex1.24x3.78 s
#3duckdb:vortex-compact1.31x4.12 s
#4datafusion:vortex-compact1.46x4.26 s
#5datafusion:parquet1.98x5.90 s
#6duckdb:duckdb2.03x5.10 s
#7duckdb:parquet2.06x5.64 s
#8datafusion:lance17.36x48.34 s
Geomean of query time ratio to fastest (lower is better)
TPC-H (NVMe) (SF=1)22 chartsDuckDB 1.4× · DataFusion 1.8× faster than Parquet

Performance Summary

#1duckdb:duckdb1.13x348.14 ms
#2duckdb:vortex1.30x383.25 ms
#3duckdb:vortex-compact1.45x419.60 ms
#4datafusion:vortex1.70x524.87 ms
#5datafusion:vortex-compact1.80x546.67 ms
#6duckdb:parquet1.87x545.51 ms
#7datafusion:parquet2.98x858.66 ms
#8datafusion:arrow3.30x1.08 s
#9datafusion:lance6.42x2.03 s
Geomean of query time ratio to fastest (lower is better)
TPC-H (S3) (SF=1)22 chartsDuckDB 1.1× faster · DataFusion 1.2× slower vs Parquet

Performance Summary

#1datafusion:parquet1.14x5.30 s
#2datafusion:vortex-compact1.31x6.00 s
#3datafusion:vortex1.32x6.23 s
#4duckdb:vortex-compact1.55x7.02 s
#5duckdb:vortex1.57x7.15 s
#6datafusion:lance1.59x7.69 s
#7duckdb:parquet1.66x7.80 s
Geomean of query time ratio to fastest (lower is better)
TPC-H (NVMe) (SF=10)22 chartsDuckDB 1.5× · DataFusion 1.6× faster than Parquet

Performance Summary

#1duckdb:duckdb1.18x1.09 s
#2duckdb:vortex1.26x1.17 s
#3duckdb:vortex-compact1.42x1.30 s
#4datafusion:vortex1.68x1.79 s
#5datafusion:vortex-compact1.87x1.90 s
#6duckdb:parquet1.95x1.75 s
#7datafusion:parquet2.76x2.59 s
#8datafusion:arrow13.90x16.39 s
#9datafusion:lance23.38x23.58 s
Geomean of query time ratio to fastest (lower is better)
TPC-H (S3) (SF=10)22 chartsDuckDB 1.6× slower · DataFusion ≈ same vs Parquet

Performance Summary

#1datafusion:vortex-compact1.19x11.44 s
#2datafusion:vortex1.19x11.23 s
#3datafusion:parquet1.19x11.73 s
#4duckdb:parquet1.30x12.43 s
#5duckdb:vortex-compact2.00x19.51 s
#6duckdb:vortex2.06x20.03 s
#7datafusion:lance3.79x43.02 s
Geomean of query time ratio to fastest (lower is better)
TPC-H (NVMe) (SF=100)22 chartsDuckDB 1.2× · DataFusion 1.8× faster than Parquet

Performance Summary

#1duckdb:vortex1.14x7.36 s
#2duckdb:parquet1.35x8.55 s
#3datafusion:vortex1.53x12.56 s
#4datafusion:parquet2.77x19.04 s
Geomean of query time ratio to fastest (lower is better)
TPC-H (S3) (SF=100)22 chartsDuckDB 3.8× slower · DataFusion 1.1× faster vs Parquet

Performance Summary

#1duckdb:parquet1.06x34.67 s
#2datafusion:vortex1.30x43.77 s
#3datafusion:parquet1.45x49.49 s
#4duckdb:vortex4.06x152.63 s
Geomean of query time ratio to fastest (lower is better)
TPC-DS (NVMe) (SF=1)99 chartsDuckDB 1.3× · DataFusion 1.5× faster than Parquet

Performance Summary

#1duckdb:duckdb1.02x1.66 s
#2duckdb:vortex1.38x2.12 s
#3duckdb:vortex-compact1.56x2.38 s
#4duckdb:parquet1.79x2.80 s
#5datafusion:vortex2.20x4.16 s
#6datafusion:vortex-compact2.44x4.70 s
#7datafusion:parquet3.24x5.93 s
#8datafusion:lance6.08x20.11 s (partial)
Geomean of query time ratio to fastest (lower is better)
TPC-DS (NVMe) (SF=10)99 chartsDuckDB 1.4× · DataFusion 1.6× faster than Parquet

Performance Summary

#1duckdb:vortex1.05x5.15 s
#2datafusion:vortex1.40x9.76 s
#3duckdb:parquet1.46x6.92 s
#4datafusion:parquet2.21x14.13 s
Geomean of query time ratio to fastest (lower is better)
TPC-DS (NVMe) (SF=100)99 chartsDuckDB 1.1× · DataFusion 1.8× faster than Parquet

Performance Summary

#1duckdb:vortex1.15x23.74 s
#2duckdb:parquet1.32x27.18 s
#3datafusion:vortex1.46x122.11 s
#4datafusion:parquet2.63x148.03 s
Geomean of query time ratio to fastest (lower is better)
TPC-DS (S3) (SF=10)99 chartsDuckDB ≈ same · DataFusion 1.2× slower vs Parquet

Performance Summary

#1datafusion:parquet1.15x48.13 s
#2datafusion:vortex1.42x55.68 s
#3duckdb:parquet1.58x55.09 s
#4duckdb:vortex1.64x60.90 s
Geomean of query time ratio to fastest (lower is better)
TPC-DS (S3) (SF=100)99 chartsDuckDB 2.1× slower · DataFusion ≈ same vs Parquet

Performance Summary

#1duckdb:parquet1.22x117.95 s
#2datafusion:parquet1.35x230.40 s
#3datafusion:vortex1.37x225.49 s
#4duckdb:vortex2.53x292.37 s
Geomean of query time ratio to fastest (lower is better)
Statistical and Population Genetics11 chartsDuckDB 2.2× · DataFusion 2.1× faster than Parquet

Performance Summary

#1datafusion:vortex1.23x3.25 s
#2datafusion:vortex-compact1.40x3.76 s
#3datafusion:parquet2.55x4.15 s
#4duckdb:vortex3.04x7.99 s
#5duckdb:vortex-compact3.40x8.56 s
#6duckdb:parquet6.67x7.54 s
Geomean of query time ratio to fastest (lower is better)
Polar Signals Profiling13 chartsVortex only

Performance Summary

#1datafusion:vortex1.00x526.52 ms
Geomean of query time ratio to fastest (lower is better)
fineweb [nvme]9 chartsDuckDB 6.2× · DataFusion 4.1× faster than Parquet

Performance Summary

#1duckdb:vortex1.01x151.37 ms
#2datafusion:vortex1.26x161.04 ms
#3duckdb:vortex-compact4.83x1.13 s
#4datafusion:vortex-compact5.08x826.45 ms
#5datafusion:parquet5.19x624.77 ms
#6duckdb:parquet6.28x799.10 ms
Geomean of query time ratio to fastest (lower is better)
fineweb [s3]9 chartsDuckDB 1.3× · DataFusion 2.3× faster than Parquet

Performance Summary

#1datafusion:vortex1.08x3.36 s
#2datafusion:vortex-compact1.17x3.52 s
#3duckdb:vortex1.61x5.28 s
#4duckdb:vortex-compact1.75x5.68 s
#5duckdb:parquet2.08x5.27 s
#6datafusion:parquet2.54x6.15 s
Geomean of query time ratio to fastest (lower is better)
appian [nvme]8 chartsDuckDB 1.1× slower · DataFusion 1.3× faster vs Parquet

Performance Summary

#1duckdb:duckdb1.18x1.69 s
#2duckdb:parquet1.18x1.66 s
#3duckdb:vortex1.25x1.83 s
#4duckdb:vortex-compact1.27x1.85 s
#5datafusion:vortex1.81x11.02 s
#6datafusion:vortex-compact1.82x11.03 s
#7datafusion:parquet2.26x11.39 s
Geomean of query time ratio to fastest (lower is better)
clickbench-sorted [nvme]10 chartsDuckDB 2.0× · DataFusion 3.2× faster than Parquet

Performance Summary

#1duckdb:vortex1.09x210.12 ms
#2datafusion:vortex1.10x230.71 ms
#3duckdb:vortex-compact1.42x274.40 ms
#4datafusion:vortex-compact1.63x397.31 ms
#5duckdb:duckdb2.12x356.55 ms
#6duckdb:parquet2.19x444.23 ms
#7datafusion:parquet3.47x1.26 s
#8datafusion:lance74.62x19.70 s
Geomean of query time ratio to fastest (lower is better)
job [nvme]113 chartsDuckDB 1.4× faster than Parquet

Performance Summary

#1duckdb:duckdb1.13x5.55 s
#2duckdb:vortex1.17x5.57 s
#3duckdb:parquet1.67x7.79 s
Geomean of query time ratio to fastest (lower is better)
tpch-nested/list sf=1 [nvme]22 chartsDuckDB 1.9× · DataFusion 1.8× faster than Parquet

Performance Summary

#1duckdb:vortex1.11x717.16 ms
#2duckdb:duckdb1.22x787.27 ms
#3duckdb:vortex-compact1.43x918.38 ms
#4datafusion:vortex1.52x1.02 s
#5datafusion:vortex-compact1.77x1.16 s
#6duckdb:parquet2.12x1.33 s
#7datafusion:parquet2.75x1.77 s
#8datafusion:lance5.74x5.11 s
Geomean of query time ratio to fastest (lower is better)
tpch-nested/list sf=100 [nvme]22 chartsDuckDB 1.2× · DataFusion 1.8× faster than Parquet

Performance Summary

#1duckdb:vortex1.07x26.05 s
#2duckdb:parquet1.33x32.07 s
#3datafusion:vortex1.43x41.37 s
#4datafusion:parquet2.52x66.47 s
Geomean of query time ratio to fastest (lower is better)
tpch-nested/list sf=100 [s3]22 chartsDuckDB 3.1× slower · DataFusion 1.2× faster vs Parquet

Performance Summary

#1datafusion:vortex1.05x117.74 s
#2datafusion:parquet1.25x136.47 s
#3duckdb:parquet1.30x137.01 s
#4duckdb:vortex4.07x469.98 s
Geomean of query time ratio to fastest (lower is better)
vortex [nvme]3 chartsDuckDB 1.8× faster · DataFusion 1.4× slower vs Parquet

Performance Summary

#1duckdb:vortex1.15x467.02 ms
#2datafusion:parquet1.19x554.84 ms
#3datafusion:vortex1.66x699.73 ms
#4duckdb:parquet2.05x832.21 ms
Geomean of query time ratio to fastest (lower is better)
SQL Benchmark File Sizes16 charts9% larger than Parquet
#1parquet
#2vortex
#3vortex-compact, measured on 4 of 6 benchmarks
#4parquet-snappy, measured on 1 of 6 benchmarks
#5lance, measured on 3 of 6 benchmarks
#6arrow-ipc, measured on 1 of 6 benchmarks
vs parquetMinGeoMeanMax
parquet: minimum 1x, geometric mean 1x, maximum 1x
vortex: minimum 0.801x, geometric mean 1.09x, maximum 2.41x
vortex-compact: minimum 0.717x, geometric mean 0.945x, maximum 1.53x
parquet-snappy: minimum 1.34x, geometric mean 1.34x, maximum 1.34x
lance: minimum 2.02x, geometric mean 2.62x, maximum 4.26x
arrow-ipc: minimum 6.24x, geometric mean 6.24x, maximum 6.24x
Geometric means of file sizes versus parquet (lower is better), over the SQL benchmarks that write Parquet with bench’s settings, each benchmark once
String Encoding Speed24 charts

Write Throughput

#1fsst1.00x224 MB/s
#2onpair-120.65x157 MB/s
#3onpair-160.64x154 MB/s

Scan Throughput

#1onpair-161.24x4.57 GB/s
#2onpair-121.24x4.62 GB/s
#3fsst1.00x3.70 GB/s
Geomean throughput ratio to FSST | Aggregate throughput (higher is better)
String Encoding Size24 charts
TotalMBvs Arrow
#1onpair-16Total: 245.4 MB2.82x
#2onpair-12Total: 243.8 MB2.82x
#3fsstTotal: 322.3 MB2.13x
vs FSSTMinGeoMeanMax
onpair-16: minimum 0.607x, geometric mean 0.755x, maximum 0.9x
onpair-12: minimum 0.6x, geometric mean 0.756x, maximum 0.877x
fsst: minimum 1x, geometric mean 1x, maximum 1x
Total file size, and geometric means of file sizes versus Arrow (higher is better) and versus FSST (lower is better)