2

I read that the implementation makes certain assumptions like the write optimization, mostly append writes, sequential deletes, but what is there in it's implementation that makes a time series database better suited for storing a time series than a mysql with proper indexing and relational structure.

perennial_
  • 1,798
  • 2
  • 26
  • 41
Sandeep Kaul
  • 2,957
  • 2
  • 20
  • 36
  • 1
    Possible duplicate of [Difference between time-series database and relational database](http://stackoverflow.com/questions/35428606/difference-between-time-series-database-and-relational-database) – Sergei Rodionov Feb 29 '16 at 21:51

2 Answers2

1

Time series databases store observations in a highly optimized format. This means faster reads, faster writes along with flexible schema. For instance, compare storage requirements in Axibase Time Series Database (my affiliation) vs MySQL for sensor readings. 7 bytes ATSD vs 63 bytes MySQL. http://axibase.com/products/axibase-time-series-database/data-storage-tests/atsd-vs-mysql/

Sergei Rodionov
  • 4,079
  • 6
  • 27
  • 44
0

The question is very generic, but here's a fairly detailed list of the tradeoffs that InfluxDB makes, and why they allow for greater performance on time-series data.

https://docs.influxdata.com/influxdb/v1.7/concepts/insights_tradeoffs/

Even tweaked to the edge of maintainability, can you push 1 million records per second into MySQL for weeks on end without it falling over? How can you bulk delete the old data that's no longer relevant while maintaining that ingest rate?

Prashant Ghimire
  • 4,890
  • 3
  • 35
  • 46
beckettsean
  • 1,826
  • 11
  • 7
  • `PARTITION BY RANGE(TO_DAYS(...))` and `DROP PARTITION` (instantly). [More discussion](http://localhost/misc/partitionmaint). I doubt if 1M rows/sec is realistic for MySQL ingestion. But think about how big your disk would need to be for even on day's data! – Rick James Mar 06 '16 at 03:40