Table 1. Benchmarking timings of the algorithm for different real-world scenarios.
Discrepancies between ‘Execution time’ and ‘Time spent in job’ can be attributed to parallelization overhead.
| Measure | Laptop (1 GB s−1) | Laptop HDD (1.6 GB s−1) | Workstation (1 GB s−1) | Workstation (RAM-disk) |
|---|---|---|---|---|
| Total time (s) | 2124 | 1648 | 1772 | 402 |
| Indexing time (s) | 227 | 201 | 229 | 228 |
| Execution time (s) | 1897 | 1446 | 1510 | 166 |
| Time spent in job (s) | 1384 | 1197.5 | 1438.03 | 160.16 |
| Loading and decompression time per worker (%) | 77.35 | 61.36 | 96.52 | 49.91 |
| Data processing time per worker (%) | 14.73 | 30.84 | 1.69 | 44.14 |
| Integration time per worker (%) | 4.30 | 6.35 | 0.26 | 5.63 |
| Save time per worker (%) | 0.02 | 1.09 | 0.01 | 0.21 |
| Overhead in job (%) | 3.60 | 0.36 | 1.52 | 0.11 |