I read Martin Heinz’s 2023 blog post “You Don’t Need a Dedicated Cache Service – PostgreSQL as a Cache” from start to finish. #DevOps or not, I tested the implementation to see what it gave back.

The dark side of a passion for technology is the risk of adopting ill‑founded beliefs about topics you don’t really master.
For example, I read a post on Hackernews about the PostgreSQL database system. Apparently PostgreSQL can also act as a cache subsystem (think Redis) and even as a message broker. Win‑win: you get rid of Redis and RabbitMQ and treat yourself to a four‑day workweek.

And PostgreSQL can even tell me when the next bus leaves for Montréal, right?

what are the requirements?

Before annoying everyone with the latest revelation picked up between two mouthfuls of peanut‑butter toast, we need to check if it actually works.

Let's open our eyes and ears and figure out what a cache is and what it requires. A cache like Redis is a system that temporarily stores data in memory so it can be retrieved much faster than if we had to go fetch it from the primary database.

So it’s a bit rich to say we’re going to use a database to cache items.
Still, we note that cached items are stored temporarily and the primary quality of a cache subsystem is read performance.

Then for the “temporary” aspect, we need a mechanism that frees items from memory after some time.

the truth is in the blog

The trick is to stretch breakfast by eating a banana or some berries. That way you can go further than the catchy headline of a blog post and learn more.

I read Martin Heinz’s 2023 blog post “You Don’t Need a Dedicated Cache Service – PostgreSQL as a Cache” from start to finish. #DevOps or not, I tested the implementation to see what it gave back.

CREATE UNLOGGED TABLE cache (
    id serial PRIMARY KEY,
    key text UNIQUE NOT NULL,
    value jsonb,
    inserted_at timestamp);

First, it’s not plug‑and‑play. The first step is pretty simple. Create an SQL table called cache and mark it UNLOGGED. An UNLOGGED table doesn’t write its updates to the WAL (write‑ahead log), which makes it faster but means it can lose its contents if the server crashes.

Not too scary, right?

Next, to make sure items don’t bloat the cache table, we need to implement some sort of TTL mechanism.
Martin Heinz suggests using the pg_cron extension to schedule jobs that will call a stored procedure to delete every item older than the given parameter.

Just the words “stored procedure” give me the same feeling as Proust’s madeleine.

In 2026, imagine your least grumpy DevOps engineer showing the new wave of developers how fun a stored procedures in a SQL database can become in the long term.

Putting toothpaste back in the tube is harder than landing a rocket vertically, isn’t it?

SWOT on Martin Heinz’s proposal

strengths

Your DevOps team is barely keeping its head above water. You already have PostgreSQL servers on hand. In less than a morning they can spin up a cache table with TTL logic and hand it over to developers.

No need to provision new resources or commit themselves to support them.

weaknesses

It’s not Redis. Developers will quickly notice they have to talk to the “cache” through their favorite ORM (or pure SQL transactions).

And they won’t be able to add “Redis” to their résumé.

Also, until proven otherwise, it’ll probably be slower than Redis.

opportunities

With the arrival of the Cloud and managed services, DBAs have vanished. By coaxing your DevOps folks into tinkering with PGSQL, you might sneak in the responsibility for administering the database servers without anyone noticing!

risks

If you know your system is heavily used, needs rock‑solid realtime performance and must scale easily, then don’t let the siren songs of the web seduce you—pick Redis.

Still sounds cool, I want to try

• Docker Desktop
• git
• DBeaver or any IDE that can connect to a pgsql server

If you have those two items (and cannot count), clone my GitHub repo sandboxes and dig in the pgsql-cache directory. Open the README.md file in Preview mode (if you’re not masochistic) and follow the instructions.

I’ve put together a little lab that will provision and configure a PostgreSQL database with a cache table and cron. Then you can run a small C# API that shows how to interact with the pgsql cache.

translated from french by Chatgpt

sources

You Don’t Need a Dedicated Cache Service - PostgreSQL as a Cache

pgsql database used as a cache