Add UUID examples from blog post

uuid_experiments/README.md

@@ -0,0 +1,16 @@
+# UUID Experiments
+
+- UUID v7 extension
+<https://pgxn.org/dist/pg_uuidv7/>
+
+- enable pg_prewarm
+
+```sh
+# needs libpq, make sure it's using Postgres.app version
+make
+make install
+create extension pg_uuidv7;
+```
+
+- Reproduce results here:
+<https://www.cybertec-postgresql.com/en/unexpected-downsides-of-uuid-keys-in-postgresql/>

uuid_experiments/create_indexes.sql

@@ -0,0 +1,19 @@
+\timing on
+
+create index on records (id);      -- records_id_idx
+create index on records (uuid_v4); -- records_uuid_v4_idx
+create index on records (uuid_v7); -- records_uuid_v7_idx
+vacuum analyze records;
+
+SELECT
+    relname,
+    pg_size_pretty(pg_relation_size(oid)),
+    pg_prewarm (oid)
+FROM
+    pg_class
+WHERE
+    relname LIKE 'records%';
+
+
+EXPLAIN (BUFFERS, ANALYZE, TIMING OFF) SELECT COUNT(id) FROM records;
+EXPLAIN (BUFFERS, ANALYZE, TIMING OFF) SELECT COUNT(uuid_v4) FROM records;

uuid_experiments/create_tables.sql

@@ -0,0 +1,31 @@
+CREATE TABLE records (
+    id int8 NOT NULL,
+    uuid_v4 uuid NOT NULL,
+    uuid_v7 uuid NOT NULL,
+    filler text
+);
+
+-- 10 million inserts
+INSERT INTO records
+SELECT
+    id,
+    gen_random_uuid(),
+    uuid_generate_v7(),
+    repeat(' ', 100)
+FROM
+    generate_series(1, 10000000) id;
+
+-- 1 million updates
+CREATE TEMP TABLE update_ids AS
+SELECT id
+FROM generate_series(1, 10000000) id
+ORDER BY random()
+LIMIT 1000000;
+
+-- Perform the updates on uuid_v4 and uuid_v7 columns (and filler text)
+UPDATE records
+SET
+  uuid_v4 = gen_random_uuid(),
+  uuid_v7 = uuid_generate_v7(),
+  filler = repeat('x', 100)
+WHERE id IN (SELECT id FROM update_ids);

uuid_experiments/leaf_pages.sql

@@ -0,0 +1,48 @@
+-- number of leaf pages
+WITH index_info AS (
+  SELECT 'records_id_idx'::text AS idxname
+  UNION ALL
+  SELECT 'records_uuid_v4_idx'
+  UNION ALL
+  SELECT 'records_uuid_v7_idx'
+),
+page_counts AS (
+  SELECT
+    idxname,
+    pg_relation_size(idxname) / 8192 AS num_pages
+  FROM index_info
+),
+all_pages AS (
+  SELECT
+    idxname,
+    generate_series(1, num_pages - 1) AS blkno
+  FROM page_counts
+),
+page_stats AS (
+  SELECT
+    idxname,
+    (bt_page_stats(idxname, blkno)).* 
+  FROM all_pages
+)
+SELECT
+  idxname,
+  COUNT(*) FILTER (WHERE type = 'l') AS leaf_pages,
+  COUNT(*) FILTER (WHERE type = 'd') AS internal_pages
+FROM page_stats
+GROUP BY idxname
+ORDER BY idxname;
+
+--        idxname       | leaf_pages | internal_pages
+-- ---------------------+------------+----------------
+--  records_id_idx      |      27323 |              0
+--  records_uuid_v4_idx |      38315 |              0
+--  records_uuid_v7_idx |      38315 |              0
+
+
+-- AFTER updates
+--  idxname       | leaf_pages | internal_pages
+-- ---------------------+------------+----------------
+--  records_id_idx      |      27589 |              0
+--  records_uuid_v4_idx |      48093 |              0
+--  records_uuid_v7_idx |      42146 |              0
+-- (3 rows)

uuid_experiments/page_density.sql

@@ -0,0 +1,56 @@
+create extension if not exists pageinspect;
+
+-- average leaf density
+WITH index_info AS (
+  SELECT 'records_id_idx'::text AS idxname
+  UNION ALL
+  SELECT 'records_uuid_v4_idx'::text
+  UNION ALL
+  SELECT 'records_uuid_v7_idx'::text
+),
+page_counts AS (
+  SELECT
+    idxname,
+    pg_relation_size(idxname) / 8192 AS num_pages  -- number of 8KB pages
+  FROM index_info
+),
+all_pages AS (
+  SELECT
+    idxname,
+    generate_series(1, num_pages - 1) AS blkno
+  FROM page_counts
+),
+page_stats AS (
+  SELECT
+    idxname,
+    (bt_page_stats(idxname, blkno)).*
+  FROM all_pages
+),
+leaf_pages AS (
+  SELECT
+    idxname,
+    100 - (free_size::float / 8192 * 100) AS fill_percent
+  FROM page_stats
+  WHERE type = 'l'
+)
+SELECT
+  idxname,
+  ROUND(AVG(fill_percent)::numeric, 2) AS avg_leaf_fill_percent
+FROM leaf_pages
+GROUP BY idxname;
+
+-- After updates
+--  idxname       | avg_leaf_fill_percent
+-- ---------------------+-----------------------
+--  records_id_idx      |                 97.64
+--  records_uuid_v4_idx |                 79.06
+--  records_uuid_v7_idx |                 90.09
+-- (3 rows)
+--
+-- v4
+-- ⚠️ Implication:
+--
+-- Write amplification is high
+-- Index bloat will grow steadily over time
+-- You will eventually need a REINDEX or pg_repack to reclaim wasted space
+-- Poorer cache locality and lookup performance may follow