diff --git a/docs/roles/debian/php-fpm.md b/docs/roles/debian/php-fpm.md
index 94a687ac4..1c6484680 100644
--- a/docs/roles/debian/php-fpm.md
+++ b/docs/roles/debian/php-fpm.md
@@ -20,7 +20,10 @@ php:
     pool_group: "{{ user_deploy.username }}" # if using unix socket this should be the web server user
     pm: dynamic # can also be static, see https://tideways.com/profiler/blog/an-introduction-to-php-fpm-tuning
     default_socket_timeout: 60
-    max_children: 5
+    # It is important to scale up processes on bigger servers, so that more
+    # requests can be handled. Double the number of vCPUs is a good default.
+    # Can be between 5 and 64.
+    max_children: "{{ [5, [ansible_facts.processor_vcpus * 2, 64] | min] | max }}"
     start_servers: 2
     min_spare_servers: 1
     max_spare_servers: 3
diff --git a/roles/debian/php-fpm/defaults/main.yml b/roles/debian/php-fpm/defaults/main.yml
index b6fea0c7d..bd11bd594 100644
--- a/roles/debian/php-fpm/defaults/main.yml
+++ b/roles/debian/php-fpm/defaults/main.yml
@@ -10,7 +10,10 @@ php:
     pool_group: "{{ user_deploy.username }}" # if using unix socket this should be the web server user
     pm: dynamic # can also be static, see https://tideways.com/profiler/blog/an-introduction-to-php-fpm-tuning
     default_socket_timeout: 60
-    max_children: 5
+    # It is important to scale up processes on bigger servers, so that more
+    # requests can be handled. Double the number of vCPUs is a good default.
+    # Can be between 5 and 64.
+    max_children: "{{ [5, [ansible_facts.processor_vcpus * 2, 64] | min] | max }}"
     start_servers: 2
     min_spare_servers: 1
     max_spare_servers: 3