The only way there can be that big a difference is if there is a flushing discrepancy in cofiguration.
Have you verified your settings are actually the same, particularly the ones containing "flush" and "sync" substrings. I seem to recall there was also a flush bug fixed since 5.5 that makes things slower since then but made 5.5 not entirely crash-safe. On Mon, 7 Oct 2024, 18:33 cyusedfzfb via discuss, <[email protected]> wrote: > Ok, tested just now with Server version: 10.4.34-MariaDB MariaDB Server: > > Server version: 10.4.34-MariaDB MariaDB Server > mysql::ping,0.0001 > mysql::select_version,0.1368 > mysql::select_all,0.7346 > mysql::select_cursor,0.9662 > mysql::seq_insert,4.4442 > mysql::bulk_insert,0.6946 > mysql::update,2.5573 > mysql::update_with_index,0.3629 > mysql::transaction_insert,4.7147 > mysql::indexes,0.2814 > mysql::delete,4.3666 > mysql::select_version::q/s,14616 > mysql::select_all::q/s,2723 > mysql::seq_insert::q/s,450 > mysql::update::q/s,782 > mysql::update_with_index::q/s,9248 > mysql::transaction_insert::t/s,424 > mysql::delete::q/s,458 > Total time,29.3222 s > Peak memory usage,23.82 MiB > > So to update the "key observations": > > - **UPDATE performance:** > - 5.5.68: *3,895,386* q/s > - 10.11.9: *1,418* q/s > - 10.4.34: *782* q/s > - **UPDATE with index:** > - 5.5.68: *4,105,436* q/s > - 10.11.9: *8,963* q/s > - 10.4.34: *9,248* q/s > - **DELETE performance:** > - 5.5.68: *4,436,065* q/s > - 10.11.9: *899* q/s > - 10.4.34: *458* q/s > > Hmm. 10.4.34 behaves much like 10.11.6, and very unlike the speedy > gonzalez 5.5.68 > > This is all very strange. > > On 10/7/24 16:41, Reinis Rozitis via discuss wrote: > >> How can it be that MariaDB 5.5.68 is performing several orders of > magnitude faster than 10.11.x for these specific operations? > > Hello, > > if you have means, could you test 10.4.34 which while EOL is the last > version based on 5.5/5.6 MySQL and issues like these appeared > > > > https://jira.mariadb.org/browse/MDEV-30501#comment-272740 > > https://jira.mariadb.org/browse/MDEV-29988 > > https://jira.mariadb.org/browse/MDEV-26055 > > > > Which while marked Closed, point to different issues/problems where > situation is still unclear .. > > > > I haven't tried the 10.11.x nor 11.x but none of the short-term release > branches after 10.4 have performed even close (though not that > dramatically) at least for our workload, so currently stuck on 10.4.x > > > > > > p.s. sorry for not answering your questions and possibly increasing the > load :) > > > > wbr > > rr > > > > > > > > _______________________________________________ > > discuss mailing list -- [email protected] > > To unsubscribe send an email to [email protected] > _______________________________________________ > discuss mailing list -- [email protected] > To unsubscribe send an email to [email protected] >
_______________________________________________ discuss mailing list -- [email protected] To unsubscribe send an email to [email protected]
