Benchmark overview
The benchmark exercised a federated analytical query against cloud-hosted data distributed across Asia and Europe. The reported end-to-end timings therefore reflect both query processing and remote-source/network behavior. In five optimized performance runs, DataIncisive averaged 320.2 seconds and 6.61 GiB peak RSS, compared with DuckDB's 301.0 seconds and 6.96 GiB. Results are specific to this query and environment.
| Characteristic | Reported configuration |
|---|---|
| Engines | DataIncisive (optimized) and DuckDB |
| Data geography | AWS source in Asia; other cloud resources in Europe |
| Workload | Cross-source federated analytical query |
| Output | 26,468,838 rows per successful run |
| Measurements | Wall time and peak resident set size (RSS) |
| Test machine | Apple Silicon (M1), 10-core CPU, 16 GB RAM, 512 GB SSD |
| Performance sample | DataIncisive: five optimized runs; DuckDB: five comparison runs |
| Additional correctness evidence | DataIncisive: 15/15 successful reruns across reported testing |
Performance comparison
| Metric | DataIncisive, optimized | DuckDB |
|---|---|---|
| Average wall time | 320.2s | 301.0s |
| Average peak RSS | 6.61 GiB | 6.96 GiB |
| Observed wall-time range | 315–347s across reported valid runs* | 293–307s |
| Observed peak RSS range | 5.34–7.29 GiB across reported valid runs* | 6.65–7.19 GiB |
| Output rows | 26,468,838 | 26,468,838 |
| Successful runs | 5/5 optimized performance runs; 15/15 correctness reruns overall | 5/5 |
*DataIncisive's reported ranges span all valid runs, not only the five optimized runs used to calculate the published averages. The DuckDB ranges describe its five comparison runs.
Optimization impact
A memory-management floor fix lowered DataIncisive's average peak RSS while slightly improving execution time.
| Metric | Before floor fix | After floor fix | Change |
|---|---|---|---|
| Average wall time | 324.4s | 320.2s | 1.3% faster |
| Average peak RSS | 7.43 GiB | 6.61 GiB | 11.0% lower |
Key findings
Comparable execution time
DataIncisive's average wall time was 6.4% longer than DuckDB's. DuckDB remained faster across the reported wall-time ranges.
Lower average memory
The optimized DataIncisive build recorded 5.0% lower average peak RSS. Individual runs varied, so this is not a per-run guarantee.
Successful query completion
All 15 reported DataIncisive correctness reruns and all five DuckDB comparison runs succeeded. Every reported successful run produced 26,468,838 output rows; row-content equivalence was not independently verified.
Cross-region workload
The sources spanned Asia and Europe. Network conditions and remote-source performance may affect end-to-end wall time.
What this benchmark shows
Both engines completed the query successfully and produced matching output row counts. DuckDB was faster on average; DataIncisive used less average peak memory. These findings apply to the reported workload, not all queries or environments.
Methodology and interpretation
The DataIncisive performance averages are calculated from five optimized runs. Its 15/15 success figure covers correctness reruns across the reported testing and must not be interpreted as 15 runs with the same performance profile. DuckDB averages and ranges come from five comparison runs. The DataIncisive observed ranges include additional valid runs beyond the optimized five-run set.
The benchmark operator reported an AWS source in Asia and other cloud resources in Europe; exact regions have not been specified. The measurements combine query execution with remote-source and network effects, and do not isolate engine-only processing time. Matching row counts do not establish full row-content equivalence.
Test machine: Apple Silicon (M1), 10-core CPU, 16 GB RAM, 512 GB SSD.
Reproducibility information to publish: exact cloud regions and source types, dataset and scanned-data sizes, query text, OS specifications, engine versions and configuration, memory limits, cache state, run order, network conditions, and RSS measurement method. Where available, include deterministic result checksums or an equivalent full-result validation.
DataIncisive's enterprise security and governance capabilities are outside the scope of this performance benchmark and should be evaluated separately.