On Wed, Aug 13, 2025 at 7:51 PM Peter Geoghegan <pg@bowt.ie> wrote:
> Apparently random I/O is twice as fast as sequential I/O in descending order! In
> fact, this test case creates the appearance of random I/O being at least
> slightly faster than sequential I/O for pages read in _ascending_ order!
>
> Obviously something doesn't add up here.
Minor clarification: If EXPLAIN ANALYZE is to be believed, "I/O
Timings" is in fact higher with the randomized "t" table variant of
the test case, compared to what I showed yesterday with the original
sequential "t" version of the table, exactly as expected. (When I said
"Apparently random I/O is twice as fast as sequential I/O in
descending order!", I was just joking, of course.)
It seems reasonable to suppose that the actual problem has something
to do with synchronization overhead of some kind or other. Or, perhaps
it's due to some kind of major inefficiency in the patch -- perhaps
the patch can sometimes waste many CPU cycles on who-knows-what, at
least in cases like the original/slow backwards scan case.
--
Peter Geoghegan
