Comments on: UUIDs are Popular, but Bad for Performance — Let’s Discuss

By: Pablo

Pablo — Tue, 03 Aug 2021 12:37:32 +0000

How about secondary indexes with UUID?

By: Murz

Murz — Mon, 26 Jul 2021 04:28:22 +0000

Thanks for the great research! Now popularity of the cuid (https://github.com/ericelliott/cuid) format, which declares benefits above UUID in binary search, is growing! Can you please make a short comparison with it too?

By: Alex Lemaire

Alex Lemaire — Fri, 30 Oct 2020 13:29:15 +0000

Great article Yves – appreciate the work you put into it.

By: Yves Trudeau

Yves Trudeau — Wed, 09 Sep 2020 14:52:53 +0000

I used 5.7, I have no plan for similar benchmarks on 8.0.x and I don’t expect a significant difference.

By: alexander-schranz

alexander-schranz — Fri, 04 Sep 2020 14:11:48 +0000

Which version of MySql was used for this benchmarks? If it 5 are there any benchmark for Version 8?

By: dharmaturtle

dharmaturtle — Fri, 21 Aug 2020 12:52:08 +0000

ULIDs are lexicographically sortable and may be generated client-side. https://github.com/ulid/spec Since they’re just 128bit numbers, you can easily insert them into UUID columns. I think they’re an excellent alternative to RFC 4122 UUIDs, especially as database identifiers

By: bw

bw — Mon, 20 Jan 2020 13:56:34 +0000

https://www.2ndquadrant.com/en/blog/sequential-uuid-generators/ explains a solution in PostgreSQL for using UUID as Pramary Key

By: Yves Trudeau

Yves Trudeau — Tue, 26 Nov 2019 14:51:01 +0000

In reply to mynameisausten. We use: https://wordpress.org/plugins/crayon-syntax-highlighter/

By: mynameisausten

mynameisausten — Mon, 25 Nov 2019 21:16:28 +0000

Unrelated to the article, but is the source code viewer that you used on this article an open source project?

By: Yves Trudeau

Yves Trudeau — Mon, 25 Nov 2019 15:55:47 +0000

In reply to Øystein Grøvlen. Indeed, the 8.0 uuid_to_bin function with the swap_flag at 1 is a good alternative, at least for the ordering impact.

By: Yves Trudeau

Yves Trudeau — Mon, 25 Nov 2019 15:48:25 +0000

In reply to Andy S. It is not as bad but you may end up in a situation where the primary key is not optimal. At least try not to store in hex.

By: Andy S

Andy S — Sun, 24 Nov 2019 21:01:20 +0000

Does using autoincrement numeric value as id (primary key) and an uuid column (unique key) on the same table have the same issues?

By: Øystein Grøvlen

Øystein Grøvlen — Fri, 22 Nov 2019 19:51:55 +0000

Note that in MySQL 8.0, the function UUID_TO_BIN can be used to encode a UUID into a BINARY(16) value. This function may also rearrange the parts of the UUID so that you will get monotonically increasing values. This should avoid the insert performance issues you discuss.