diff --git a/sql/blob-size-stats/.gitignore b/sql/blob-size-stats/.gitignore
new file mode 100644
index 0000000..3bedf80
--- /dev/null
+++ b/sql/blob-size-stats/.gitignore
@@ -0,0 +1 @@
+blob-size-stats.csv
diff --git a/sql/blob-size-stats/analyze.py b/sql/blob-size-stats/analyze.py
new file mode 100755
index 0000000..a58194d
--- /dev/null
+++ b/sql/blob-size-stats/analyze.py
@@ -0,0 +1,8 @@
+#!/usr/bin/env python3
+
+import pandas as pd
+
+DATA_FILE = "blob-size-stats.csv"
+
+df = pd.read_csv(DATA_FILE, low_memory=False, skiprows=1, names=["blob size"])
+print(df.describe())
diff --git a/sql/blob-size-stats/blob-size-stats.log b/sql/blob-size-stats/blob-size-stats.log
new file mode 100644
index 0000000..56a4b26
--- /dev/null
+++ b/sql/blob-size-stats/blob-size-stats.log
@@ -0,0 +1,21 @@
+$ date -R
+Sat, 30 Jan 2021 19:40:48 +0100
+
+$ ./blob-size-stats.sh 
+real	59m17,410s
+user	0m5,709s
+sys	0m3,323s
+
+$ wc -l blob-size-stats.csv 
+9737167 blob-size-stats.csv
+
+$ ./analyze.py
+          blob size
+count  9.737166e+06
+mean   8.233840e+04
+std    1.207056e+06
+min    2.000000e+00
+25%    1.054000e+03
+50%    3.313000e+03
+75%    1.308800e+04
+max    1.109894e+08
diff --git a/sql/blob-size-stats/blob-size-stats.sh b/sql/blob-size-stats/blob-size-stats.sh
new file mode 100755
index 0000000..10b13d6
--- /dev/null
+++ b/sql/blob-size-stats/blob-size-stats.sh
@@ -0,0 +1,2 @@
+#!/bin/bash
+time psql --csv service=swh-replica < blob-size-stats.sql > blob-size-stats.csv
diff --git a/sql/blob-size-stats/blob-size-stats.sql b/sql/blob-size-stats/blob-size-stats.sql
new file mode 100644
index 0000000..5d5759b
--- /dev/null
+++ b/sql/blob-size-stats/blob-size-stats.sql
@@ -0,0 +1 @@
+select length from content tablesample bernoulli (0.1);