Difference between time-series database and relational database

xtiger picture xtiger · Feb 16, 2016 · Viewed 7k times · Source

I have read some concept about time-series database and some answers about the differences between these 2 but I can't still get my head around these differences.

How data is written and stored to disk that make the differences between time-series database and rdbms?

What are the best use cases for time-series database over rdbms and VICE VERSA?. I mean the cases that make the time-series database provide outstanding performance over rdbms, also I want to know the cases that rdbms is a more suitable choice than time-series database.

Answer

Sergei Rodionov picture Sergei Rodionov · Feb 16, 2016

Time Series Database Advantages:

  1. Throughout 100K+ to 1M+ inserts per second
  2. Bytes stored per time/value tuple: 2-10 vs 30-100 (rdbms)
  3. Built-in time series transformation and aggregations functions
  4. Schema optimized for time-series arrays with built-in sharding and indexing

Relational Database Advantages:

  1. Full SQL support
  2. Ability to store any data other than time-series
  3. Extensive DBA resources and tooling

By writing your own application code and stored procedures you can accomplish almost everything that TSDBs provide, but you may end up with an implementation that is slower and more demanding in terms of underlying compute and storage resources.

For all practical purposes and with some compromises to referential integrity, you can have both databases run concurrently: store your extended app schema in relational database and time-series in TSDB.