Support for Postgres extension TimescaleDB?


#1

The current recommendation seems to be to put time series metadata in Postgres and the measurements themselves in Cassandra for performance reasons. Is there any plan to support the Postgres extension TimescaleDB for storing the measurement data? It seems to be gaining traction as a time series storage engine that supports full SQL queries in Postgres.

More info on TimescaleDB here: https://www.nextplatform.com/2018/01/25/time-time-series-databases/


#2

TL;DR lets wait to see it get to v2 and have a production use case (today their version is v0.8… and their db was announced < 30 days ago)

Explanation:
We stopped looking into beta and v1 version of databases because:
a) it takes significant efforts to validate performance and more importantly stability, consistency and cost (and so far most of them failed in one of the areas)
b) this is not where current PSR and cost bottlenecks are in the platform

We do keep a list of projects / companies to follow and I will add this one to it


#3

Another datapoint to consider regarding TimescaleDB:


#4

Latest news:


#5

@benj Thanks for the post. We’ll definitely look into it. Just curious, What are the current characteristics of your data (size & throughput), specifically:

  1. How many sensors / tags are you looking to ingest / analyze data for?
  2. What is the granularity of the incoming data?
  3. What throughput is the data coming int at?
  4. What is the expected output time for insight for this data?

Given the above characteristics, what is the current bottleneck in the platform that is preventing us from achieving the business goal?

While we’re continuously enhancing the time series infrastructure at C3 (and would love to investigate timescale db), we would love to know what use cases which cannot be solved by the C3 time series engine, that are triggering the need for a specific solution?

We definitely want to build the next generation time series experience on the c3 IoT platform (probably even much better than timescale db). Would love to understand about the specifics.


#6

Thousands of measurements per asset at hundreds of samples per second. Ask Annaji Garmiella for more information.