4 ms·

Some more thoughts on random_page_cost

• nopurpose 10d ago
  > a sequential scan may push ~100GB of other data from memory (shared buffers or page cache). Small note, if I remember correctly PostgreSQL allocates small buffer within shared_buffers for bulk reads. It is so that single sequential scan won't flush all of shared buffers content. Similar mechanism exists for a pagecache implementation on Linux. It is divided into generations based on access patterns.