Benchmarking Results Different number and type of GROUP BY attributes

Jump to bottom Edit New page

Alexander Löser edited this page Feb 5, 2019 · 7 revisions

performed in hyriseConsole on custom generated tables with 1/2/3 int/string column, depending on the respective benchmark, dictionary encoding
always exactly one aggregate: SUM(agg), where agg is a int column with random values ranging from 1 to 10
values for group by columns have the same random value range
this leads to up to 10, 100, 1000 possible different group by groups

AggregateSort

1i	255 ms 995 µs
2i	553 ms 867 µs
3i	845 ms 280 µs
5i      1 s 253 ms
7i      1 s 858 ms
1s	948 ms 804 µs
2s	1 s 802 ms
3s	2 s 547 ms
5s      3 s 619 ms
7s      5 s 990 ms

AggregateHash

1i	68 ms 527 µs
2i	113 ms 453 µs
3i	242 ms 221 µs
5i	360 ms 274 µs
7i      395 ms 476 µs
1s	97 ms 807 µs
2s	176 ms 559 µs
3s	328 ms 605 µs
5s      486 ms 282 µs
7s      645 ms 439 µs