Hot and cold cache behavior

Returning to the regular version of the query seen previously, that it executed in 7.994 milliseconds. This represents "hot" cache behavior, meaning that the data needed for the query was already in either the database or operating system caches. It was left behind in cache from when the data was loaded in the first place. Whether your cache is hot or cold (not in the cache) is another thing to be very careful of. If you run a query twice with two different approaches, the second will likely be much faster simply because of caching, regardless of whether the plan was better or worse.

You can look at how long a query against the entire table takes as a way to measure the effective transfer rate for that table. In the hot cache case, it gives you an idea how fast data moves between two sections of memory:

SELECT pg_size_pretty(CAST(pg_relation_size('customers') / 7.994 * 1000 as int8)) AS bytes_per_second;
 bytes_per_second 
----------...