Stephan Garland stephanGarland

Environment

System

Macbook Pro M3
RAM: 36 GiB
SSD: 500 GB

MySQL

Results

All programs compiled with -O3, using clang v15.0.0 on MacOS, and gcc 10.2.1 on Linux. Mac was an M1 Air base model (8 GiB RAM), with low power mode disabled. Linux server was a Debian 11 VM with a Xeon E5-2650 v2 @ 2.60GHz, and 64 GiB of PC3-12800R RAM.

Linux

❯ ./uuid_runner.sh

Environment

System

OS: Debian 12 VM on Proxmox VE 8
CPUs: 8x Xeon E5-2620 v2
RAM: 8 GiB DDR3
HDD: Samsung PM863 NVMe in Ceph over Mellanox Infiniband 56 Gbps IPoIB
- Virtualized via scsi-virtio-single, aio=native, io_threads=enabled
FS: ext4 (rw,noatime)

Introduction

MySQL's documentation states:

When combining LIMIT row_count with DISTINCT, MySQL stops as soon as it finds row_count unique rows.

The question was asked, how is this deterministic / accurate? My hypothesis is that it's due to MySQL's use of a clustering index, as opposed to Postgres' heap storage. I think that, given a monotonic PK such as an AUTO_INCREMENT (or perhaps any index), it's able to use that to guarantee determinism.

This table has 1,000,000 rows, consisting of a UUIDv4 PK, a random int of range (1,1000000), and ~1 KiB of Lorem Ipsum text.

postgres=# \d+ uuid_pk
                                          Table "public.uuid_pk"
 Column  |  Type   | Collation | Nullable | Default | Storage  | Compression | Stats target | Description
---------+---------+-----------+----------+---------+----------+-------------+--------------+-------------
 id      | uuid    |           | not null |         | plain    |             |              |
 user_id | integer |           | not null |         | plain    |             |              |
 lorem   | text    |           | not null |         | extended |             |              |

System Specifications

OS: Debian Bullseye 5.10.0-23-amd64
Virtualized: Yes (Proxmox)
CPU: E5-2650 v2 @ 2.60GHz
Allocated Core Count: 16
Allocated RAM: 64 GiB PC3-12800R
Disk: Samsung PM983 1.92 TiB via Ceph
Filesystem: XFS
Mount Options: defaults,noatime
Postgres Version: 15.3

	import itertools
	import uuid
	from functools import lru_cache
	from typing import Dict, List, Optional, Union

	from sqlalchemy import (Column, DateTime, ForeignKey, Integer, String,
	create_engine)
	from sqlalchemy.dialects import mysql
	from sqlalchemy.orm import (Session, declarative_base, relationship,
	sessionmaker)

	SET search_path TO 'example';

	INSERT INTO genre (name) VALUES ('Rock'), ('Classic Rock'), ('Metal'), ('Progressive Metal'), ('Arena Rock'), ('Alternative Metal'), ('Grunge');
	INSERT INTO category (name) VALUES ('Book'), ('Music');
	INSERT INTO format (name) VALUES ('Compact Disc'), ('Vinyl');

	INSERT INTO band (year_formed, name) VALUES (1985, 'Majesty'), (1988, 'Dream Theater'), (1990, 'Tool'), (1970, 'Queen'), (1987, 'Alice in Chains');

	INSERT INTO artist (first_name, last_name) VALUES ('John', 'Petrucci'), ('John', 'Myung'), ('James', 'LaBrie'), ('Jordan', 'Ruddess'), ('Mike', 'Portnoy'), ('Mike', 'Mangini');
	INSERT INTO artist (first_name, last_name, prefix, suffix) VALUES ('Brian', 'May', 'Sir', 'CBE');

	#!/usr/bin/env python3

	# LICENSE
	# This Source Code Form is subject to the terms of the Mozilla Public License, v. 2.0.
	# If a copy of the MPL was not distributed with this file, You can obtain one at http://mozilla.org/MPL/2.0/.

	# Copyright 2024 Stephan Garland

	"""
	Calculates various parameters for the InnoDB buffer pool based on a simple input.

	mysql> CREATE TABLE foo(id TINYINT UNSIGNED AUTO_INCREMENT PRIMARY KEY, bar BINARY(16), baz CHAR(36));
	Query OK, 0 rows affected (0.03 sec)

	mysql> INSERT INTO foo (bar, baz) VALUES (UUID_TO_BIN('dc23a9b9-a129-11ee-95fb-0242ac110000'), 'dc23a9b9-a129-11ee-95fb-0242ac110000');
	Query OK, 1 row affected (0.02 sec)

	mysql> INSERT INTO foo (bar, baz) VALUES (UUID_TO_BIN('dc23a9b9-a129-11ee-95fb-0242ac110000'), 'dc23a9b9a12911ee95fb0242ac110000');
	Query OK, 1 row affected (0.01 sec)

	mysql> SHOW BINARY LOGS;