search

influxdata / influxdb

Scalable datastore for metrics, events, and real-time analytics
https://influxdata.com
Apache License 2.0
28.48k stars 3.53k forks source link

run very slow #880

Closed gabrieldodan closed 9 years ago

gabrieldodan commented 9 years ago

The server: 8 cores CPU 15G RAM HDD WDC WD2502ABYS-18B7A0 http://wdc.custhelp.com/app/answers/detail/a_id/2545/~/specifications-for-the-250-gb-re3-drive-(model-wd2502abys) CENTOS 6.5

I have 100 series each contain around 800 points. A query like this select field1, field2 from s1, s2, s3, ....., s100 limit 10 took around 3 seconds. That's very very slow! I have same data into a postgresql database and it run faster, it took under 1 second to fetch data from 100 tables. Is it normal to take so much time or I am doing something wrong ? Thanks.

pauldix commented 9 years ago

What version are you running? What is your max-open-files setting? Please try running on 0.8.1.

Dieterbe commented 9 years ago

possibly related to #821 , i.e. despite your limit 10 it might actually traverse all records in the series. (if you remove the limit clause, is performance similar?) if that's the case, then 3000ms for 80k records is about 0.037 ms per record, from my experience, you should be able to get it down to under 0.020ms per record. but of course if you're suffering from #821 then you'll see a big speed improvement (should be roughly 2-8x) once that's fixed.

toddboom commented 9 years ago

Assuming this will be fixed with #821, closing out.