tsbs VS orioledb

1. correctness: from small units tests to relatively complex integrations tests. they typically populate a test database and query it via various interfaces, such as REST or the Postgres protocol. we use Azure Pipelines to execute them - testing in MacoOS, Linux (both Intel and ARM) and Windows.

2. performance: we tend to use the TSBS project for most of our performance testing and profiling. fun fact: we actually had to patch it as the vanilla TSBS was a bottleneck in some tests. Sadly, the PR with the improvements is still not merged: https://github.com/timescale/tsbs/pull/186

As usual, we use the industry standard Time Series Benchmark Suite (TSBS) as the benchmark tool. Unfortunately, TSBS upstream does not support MongoDB time series collections.

hey hn, supabase ceo here

we've been fans of Oriole for a while now and have been long-time supporters

in case you're jumping straight to the comments: OrioleDB is a table storage extension for Postgres. it acts as a drop-in replacement for the default postgres storage engine using the Table Access Method APIs (pluggable storage). the storage engine changes the representation of table data on disk. its architecture is designed to take advantage of modern hardware like SSDs and NVRAM. it implements MVCC, the feature that allows allows multiple connected users to see different versions of the data depending on when their transaction started, via an UNDO log rather than tuple versioning.

one caveat: it requires several patches to the postgres core to expand on the type of features external storage engines extensions can implement. for this reason it could be a while before you see this land as a default engine on supabase. we will probably make it available as an option for customers who want to experiment - no timeline is decided yet.

finally, we have been working with the team on decoupled storage and compute [0]. this is experimental but promising, especially with some recent advances in S3 (specifically Express One Zone [1]). we have a demonstration in the blog post.

i'll message Alexander in case there are any technical questions

[0] https://github.com/orioledb/orioledb/blob/main/doc/usage.md#...

[1] https://aws.amazon.com/s3/storage-classes/express-one-zone/

When I saw "cloud native" I was expecting S3-ish the way Neon does it but they say it's experimental: https://github.com/orioledb/orioledb/blob/beta4/doc/usage.md... and for them to say "beta, don't use in production" and then a separate "experimental" label must make it really bad

There are some interesting things in development to potentially solve that problem.

Here's a recent HN submission about OrioleDB of the more promising ones: https://news.ycombinator.com/item?id=36740921

Source code: https://github.com/orioledb/orioledb

https://github.com/orioledb/orioledb/blob/main/doc/arch.md

> - PostgreSQL is very conservative (maybe extremely) conservative about data safety (mostly achieved via fsync-ing at the right times), and that propagates through the IO stack, including SSD firmware, to cause slowdowns

This is why our first goal is to become pure extension. Becoming part of PostgreSQL would require test of time.

> - MVCC is very nice for concurrent access - the Oriole doc doesn't say with what concurrency are the graphs achieved

Good catch. I've added information about VM type and concurrency to the blog post.

> - The title of the Oriole doc and its intro text center about solving VACUUM, which is of course a good goal, but I don't think they show that the "square wave" graphs they achieve for PostgreSQL are really in majority caused by VACUUM. Other benchmarks, like Percona's (https://www.percona.com/blog/evaluating-checkpointing-in-pos...) don't yield this very distinctive square wave pattern.

Yes, it's true. The square patters is because of checkpointing. The reason of improvements here is actually not VACUUM, but modification of relevant indexes only (and row-level WAL, which decreases overall IO).

I took a look at https://github.com/orioledb/orioledb which is a project attempting to remedy some of Postgres' shortcomings, including MVCC. It looks like they're doing something similar to MySQL with a redo log, as well as some other optimizations. So maybe this is the answer.

You don’t need a database written (or rewritten in Rust). we’re working to make Postgres scalable for the next decade too https://github.com/orioledb/orioledb

> I’d love to see B-Tree primary storage option. Aka store the row data inside the primary index.

It is coming: https://github.com/orioledb/orioledb

sorry to underwhelm!

if you like Neon, then I imagine you like their database branching model? On Friday we announced[0] our 500K investment into OrioleDB, who are working on branching[1], with the plan to upstream these changes into Postgres core.

It would be possible for us to run a fork of Postgres today which supports branching, but our long-term view is that developers would prefer a non-forked version of Postgre (to mitigate any risk of lock-in). So we will work on adding branching to Postgres core in the background, which will be a benefit to the entire Postgres ecosystem.

[0] Announcement:https://supabase.com/blog/supabase-series-b#where-were-going

[1] https://github.com/orioledb/orioledb/wiki/Database-branching