Stop Cramming Everything into Postgres

Loved the discussion on this hn post few months ago on Postgres for everything. https://news.ycombinator.com/item?id=42347606 I thought I'd share my thoughts over a separate post. I see a bunch of comments: just don’t do it! I’m glad this is becoming mainstream, and people are realizing that Postgres for everything doesn’t work. Don’t get me wrong—I’m a huge Postgres proponent and have spent 10 years helping customers implement it. However, I’m a strong believer in using Postgres for what it’s designed for in first-place. Postgres was designed as a row-based OLTP database, with over 30 years of effort dedicated to making it robust for that use case.

I know there are many extensions attempting to make Postgres support other use cases, such as analytics, queues, and more. Keep in mind that these extensions are relatively recent and aim to retrofit new capabilities onto a database primarily designed for transactional workloads. It’s like adding an F1 car engine to a Toyota Camry — will that work?

Extensions also have many issues—they are not fully Postgres-compatible. In Citus, for example, we added support for the COPY command 4 years into the company, and chasing SQL coverage was a daily challenge for 10 years. Unable to use the full capabilities of Postgres and having to work around numerous unsupported features defeats the purpose of being a Postgres extension. This was a common feedback across Citus customers - when you say Postgres extension, customers are looking for something that supports all Postgres features! Side note for the Postgres extension companies: Chasing Postgres compatibility and world class performance are ginormous problems and are hard to achieve without laser focus on each of them.

On the other hand, you have purpose-built alternatives like ClickHouse, Snowflake for analytics, Redis for caching, and Kafka for queues. These technologies have benefited from decades of development, laser-focused on supporting specific use cases. As a result, they are highly efficient for their intended purposes.

I often hear that these Postgres extensions are expanding the boundaries of what Postgres can do. While I partly agree, I question the extent to which these boundaries are truly being expanded. In this era of AI, where data is growing exponentially, handling scale is critical for any technology. These boundaries will likely be broken very quickly - A data point here is that in my day-to-day now, I'm seeing AI companies use purpose-built technologies like ClickHouse way sooner. A few years ago, it was a Series A or Series B company that would consider a purpose-built database, now it is Seed stage companies. This is because the amount of data / users to be handled has grown tremendously from the get go and companies are preferring solutions that scale with their workload.

Also I keep hearing these Postgres extension companies say Zero ETL in their positioning, which I don't agree. Let’s take a search or an analytics extension where you want to run analytics on your transactional data. You either need to manage a cron to continuously offload the transactional data to the storage format of the extension or if the extension supports logical replication (not all do), you need to have logical replication setup. And logical replication is ETL! It is painful to manage logical replication - fun fact, at PeerDB, Postgres was the second largest target connector as logical replication was painful to manage, at scale. And yes, you could have capabilities which does the conversion in the background, but that is again a ginormous problem, I think only a few fully fledged databases like Vertica have this feature.

TL;DR: Please think carefully before choosing the right technology as you scale. Cramming everything into Postgres might not be the best approach for scaling your business.