Please don't reply to lustre-devel. Instead, comment in Bugzilla by using the following link: https://bugzilla.lustre.org/show_bug.cgi?id=10902
Created an attachment (id=9231) Please don't reply to lustre-devel. Instead, comment in Bugzilla by using the following link: --> (https://bugzilla.lustre.org/attachment.cgi?id=9231&action=view) Quick enhancement to current lock matching code This patch is a simple speedup of lock comparison code. While nowhere as sophisticated as bug 11300 or skiplists code here (developed for b1_5 currently), this code reduces n^2/2 comparisons required to grant N identical compatible locks to only N comparisons (if there was nothing granted). It achieves such a task by assuming that if we have same mode lock already granted with policy that is a superset of requested lock policy, then there are no conflicting locks to requested lock in a granted queue and so we can return success immediatelly. Situations it should help with: many clients do ls on a same dir/file. (read on a same file) Situations that it has no effect on: After above many clients ls on same dir/file, touch dir/file1 (no easy equivalent for extent locks) _______________________________________________ Lustre-devel mailing list [email protected] https://mail.clusterfs.com/mailman/listinfo/lustre-devel
