Time Series Database Lectures #1 - Paul Dix (InfluxDB)

InfluxDB Storage Engine Internals
Paul Dix (Co-founder & CTO, InfluxData)
September 14, 2017
Annotated Video: scs.hosted.pan...
Time Series Database Lectures (Fall 2017)
Carnegie Mellon University
db.cs.cmu.edu/s...
Organized by Andy Pavlo
www.cs.cmu.edu/...

[@山中三次郎]

Hmmm, don't really understand the part between 35:19 and 36:05. Wondering anyone can elaborate more?
- Why are tag blocks and measurement blocks in the index file layout(35:19)?
- How do we query a series name in the series block (35:34)？
- Why do we need to do a hash on the series name to get the bucket (35:47)?

[@varunvats32]

1. This could be due to their support for tag grouping and filtering (on measurement as well).
2 & 3. Not sure

[@jackielii]

"We use mmap". I knew Andy was going say something :D

[@ozgurakkurt9770]

and they changed it later :)

[@ambarishhazarnis9531]

What a tough room

[@abramswee]

so which TSDB is the best at this moment in the market? currently exploring for one to store smart meters. any suggestion>?

[@Pifagorass]

Test with your data and your queries multiple engines, check how their fir to you environment (existing reporting dashboard software e.t.c). I use one with PGSQL API just to get covered in R/Python/Java and whatever clients want to connect with. In your case other aspects may be more important e.g. compression ratios or ingestion rates ...
TLTR you know better what's best for your after testing multiple.

[@mahanteshambali]

Try kdb

[@veixquadron2928]

the opening music? anyone?

[@fishdust]

Veix Quadron - the opening music is “we the people” by A Tribe Called Quest

[@allrightbruh]

A Tribe Called Quest

[@damiansteele8701]

@11:16 its tuh·
pl

[@niceperson6412]

The questions got asked in the session suck.

[@engineerv3248]

Could you upload HD or FHD video?

[@CMUDatabaseGroup]

See the annotated video on Panopto: scs.hosted.panopto.com/Panopto/Pages/Viewer.aspx?id=a2b94da7-a4a3-4c54-8bea-6b665e33da6b