How to evaluate your next database

citusdata.com

33 points·samaysharma··10 comments

Just use PostgreSQL and don't evaluate anything. PostgreSQL can handle an enormous number of use cases, scales very well and can be made redundant with read-only nodes too.

If you outgrow PostgreSQL, you have an excellent problem and are probably making tons of money. Don't waste mental power on this; just use it and focus on your business/product's core value proposition.

Or, read the article, which reaches the same conclusion with way more words. :)

For RDBMSes, this.

For write-intensive time series, or highly distributed, or graph databases you might consider other options.

For time series there is the Timescale extension for PostgreSQL.

As the sibling pointed out, there are extensions for that, that may be useful in PostgreSQL.

It also is a good KVS, a document store, a queuing system, a pub/sub system, etc. While I usually add a message bus and a memory cache to a larger system, PostgreSQL can do all these things for you if you want to keep things simple.

Is there a good extension for postgresql that turns it into a graph database?

There is AgensGraph, but it seems to be a fork, not an extension.

I think the default should be some RDBMS. Unless you already know your domain - which is unlikely. Get to know your product, problem space, domain - then with more information you can start optimising by adding more specialised db engines (if this optimisation provides benefits of course).

PG 11 + Citus + cstore_fdw + PipelineDB is my preffered Database (OLAP + OLTP + Streaming middleware + HA) stack choice

TLDR: Database vendor for Postgres makes Postgres look good.

My startup uses Postgres though and we like it.

Postgre is also my go-to relational database.

Reply on news.ycombinator.com