"group by" performance

[JasonLi-cn]

Compare "group by" performance with ClickHouse

Question:
What configuration of Databend can I change to improve performance?

CVM info:

• 84 processor (cat /proc/cpuinfo| grep "processor"| wc -l)
• Mem total: 314G

SQL:
SELECT count(*) FROM numbers_mt(10000000) group by number limit 5;

Databend:

ClickHouse:

[sundy-li]

Yes, as the number of stored keys increases we did not have the two-level hashmap optimization like ClickHouse.
It can be enhanced in the future.

[sundy-li]

It's done in #8577