Hi Katte
Katte <liu.z...@ccfan.com.cn> 于2019年9月24日周二 下午1:56写道： > Hi All, > > I have 100 millions data is stored in Doris. I testing rullup performance > as the following: > > The fist test is was normal and performace improved a lot after created > rollup. > ------------------------------------------------------ > # no create rollup --- costs 4.73 sec > SELECT key1, key2, SUM(value_10) FROM tb06 group by key1, key2; > > # add rollup > ALTER TABLE tb06 ADD ROLLUP rollup_key7(key1, key2, value_10); > > # after created rollup --- 0.02 sec > SELECT key1, key2, SUM(value_10) FROM tb06 group by key1, key2; 0.02 > ------------------------------------------------------ > > But the second test found no significant improverment in performance after > created rollup. > ------------------------------------------------------ > # no create rollup --- costs 21.75 sec > SELECT prod, region, channel, SUM(value_2) FROM tb06 group by prod, > region, channel; > # add rollup > ALTER TABLE tb06 ADD ROLLUP rollup_key9_1(prod, region, brand, value_2) > from rollup_key9; > # after created rollup --- 20.97 sec > SELECT prod, region, channel, SUM(value_2) FROM tb06 group by prod, > region, channel; > > 1. you should make sure that this query is through the rollup 2. What is the size of the base table and the rollup table? Is the rollup table smaller than the base table? 3. Is there other key column other than "prod, region, channel" 4. You can set is_report_sucess = true, and execute the query again, and get the query profile in Doris Frontend's webserver. > I want to know if this test is normal, and rollup does not work very well > when there is a large amount of data and more number of keys grouped > together. > > > > > > > 发送自 Windows 10 版邮件应用 > >
