34 comments

[ 5.0 ms ] story [ 53.9 ms ] thread
Introducing the first major release of TideSQL for MySQL. A transactional storage engine built from first principles and on modern research that's write and space optimized, highly effective on reads, and offers parity with InnoDB. Do check it out, we would love your thoughts!
My thoughts: please, please get someone to rewrite this page from LLM poetry to a real person's writing.
As for me, I've copied and pasted the text into Claude, prompted: "Distill key insights" and got everything I needed to know.

Link: https://claude.ai/share/6cf78509-e283-49cc-89b1-c892bfce0641

In fact... I read most of HN like this nowadays, including the comments.

This is really not better.
> This is really not better. Well… not much better.

That's just your unsubstantiated opinion. I'm saving time and getting the meat.

Is your own opinion more substantiated? Weird thing to say.
It is actually, I provided the artifact and stated that it saves time and provides the substance. You said nothing.
One shares his experience and how he likes to do things.

The other states his opinion as a fact.

And your asking the former why his opinion is more substantiated? Bro, he never claimed that.

You feel that way is because unhappy that AI able to do something very similar to what you can do at a fraction of effort. I have read many text by human and AI, and many time AI is superior in description though more verbose. Human writing are very creative and terse if well written. But for huge outputs of comparisons, human writing is overrated. Majority quality is nowhere near a sloppy AI level. Judge it objectively not a pre-built bias saying every AI output is bad
I think most of the market these days is running a managed database - AWS RDS, GCP CloudSQL, etc. Any news on when people could expect this to be a supported plugin?
Very much possible over time. Currently it’s not. We are working closely with MariaDB and Oracle on making TidesDB available in server releases through extended means. MariaDB you can already use TidesDB from version 11, included through their Foundry.
How does this differ from RocksDB / MyRocks?
MyRocks have been archived for MySQL. MyRocks in MariaDB is using a very old version of RocksDB in which is rather buggy. TideSQL is similar to MyRocks but it goes a lot further with it's features, trying to offer parity with InnoDB. TideSQL's core is using TidesDB the library which is a write, and space optimized storage engine written in C from first principles, so from the ground up.
So if Meta stopped maintaining MyRocks, are they moving to a different solution?
Not sure at the moment what they are moving to or have moved to. I just heard from Mark Callaghan few months ago about the archiving, not much besides that.
My understanding is that Meta continues to use MySQL and specifically MySQL, but have decided they aren't investing in upgrading to the newer MySQL versions. And possibly their fork contains a lot of custom patches beyond just MyRocks. So as a consequence of this decision they no longer support MyRocks for newer versions of MySQL and otoh MySQL no longer supports the 5.x versions.
Curious to know if MySQL has any distinct advantages over Postgres at this point.
Not needing a connection pooler (MySQL scales to thousands of connections via threads without breaking a sweat) and not having to deal with garbage collection (vacuuming) are two.
I love PostgreSQL, but the clustering story is definitely easier in MySQL land.
Finally a replacement for RocksDB / MyRocks. I want to move from MySQL to Postgres, but without a space optimized storage as the MySQL has in Percona, it is impossible to me.
Indeed! Postgres vacuum can be troublesome on lots of write and delete heavy workloads and no way to just set a table to another engine. TidesDB is highly efficient on amplification rather that be write, read, or space for an LSM based engine. Reads are also pretty great with our LSMB+tree hybrid architecture; You can read about TPC-C analysis against InnoDB in MariaDB (larger than memory) and see TidesDB keeping up and even surpassing InnoDB on this very heavy benchmark.

https://tidesdb.com/articles/benchmark-analysis-hammerdb-tpc...

Is there some specific reason/feature why you would prefer Postgres? Or you just prefer it as a standard or some non-technical reasons? (The latter is valid too f course, just more subjective.)
I’d say Postgres is generally very fast for the average workload in analysis and the external extensibility is truly great, being able to use different index types for example. Postgres also has good support for modern SQL and many datatypes. With those 2 aspects the ecosystem is rather large thus lots of ways to extend Postgres. MySQL and MariaDB are now starting to expand their horizons in this regard.
Is there also a comparison with innodb with compressed tables (Storage size/speed/cpu load/mem usage )? (I assume at least some of the benefit in storage/performance may be due to the compression being on by default)
For MySQL yes on by default, for MariaDB no off by default. In this articles analysis compression was on indeed.
For the analysis comparison question, yes many articles but because InnoDB does sync compression and not async like TidesDB it’s not performant, but there is analysis on our blog, and against RocksDB as well at library and server level.