[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

[Bug 221152] GFS2: Clean up of glock code



Please do not reply directly to this email. All additional
comments should be made in the comments box of this bug.


https://bugzilla.redhat.com/show_bug.cgi?id=221152





--- Comment #18 from Steve Whitehouse <swhiteho redhat com>  2008-11-20 10:11:09 EDT ---
With the latest patch for eliminating glockd and scand, I have had another idea
to help try and improve the performance of glocks. Ideally we'd like to be able
to solve the writeback problem and also that of efficient scheduling of inode
reads (via inode_go_lock at the moment).

It occurred to me that if we introduced a new concept, that of an elevator
based workqueue, we could solve both problems at the same time. Each entry in
the workqueue would have an index, which in this case would be determined by
the inode number in question. This, in conjunction with an elevator algorithm
when the workqueue is being run, should result in I/O being issued in a more
optimal order. It also means that we can then deal with both reads and writes
with a single solution since both writeback and inode reads are started under
the workqueue.

>From comment #16, items 7-9 are done in the latest patch. From comment #17,
items 12 & 13 are done and item 14 is high on the list. There is still a lot of
work to do in this area though.

-- 
Configure bugmail: https://bugzilla.redhat.com/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]