Skip to content

The first execution of a query is usually faster than subsequent ones, even after benchmark restarts #15527

@SloNN

Description

@SloNN

Running Clickbench benchmark.

The database running on current main version. The database is configured with 4 dynamic nodes + 1 storage node on each of the three servers. Each node (static and dynamic) gets 100GB of memory. Every node reports memory usage about 1 GB.

Query 22 (and a number of others, less noticeable there) behave unexpectedly in the first run:

Query22:
 iteration 0: ok 7.283150s seconds
 iteration 1: ok 0.579033s seconds
 iteration 2: ok 0.518693s seconds

<restart test>

Query22:
 iteration 0: ok 6.790968s seconds
 iteration 1: ok 0.606593s seconds
 iteration 2: ok 0.577245s seconds

<restart test>

Query22:
 iteration 0: ok 4.007444s seconds
 iteration 1: ok 0.510473s seconds
 iteration 2: ok 0.512680s seconds

Query22:
 iteration 0: ok 5.038643s seconds
 iteration 1: ok 0.525281s seconds
 iteration 2: ok 0.516553s seconds
 iteration 3: ok 0.505478s seconds
 iteration 4: ok 0.564580s seconds
 iteration 5: ok 0.491766s seconds
 iteration 6: ok 0.500325s seconds
 iteration 7: ok 0.493554s seconds
 iteration 8: ok 0.505824s seconds
 iteration 9: ok 0.501760s seconds 

Metadata

Metadata

Assignees

Type

No type

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions