[Ocfs2-users] Concurrent write performance issues with OCFS2

[Erik Schwartz]

Thanks for your reply.


On 2/28/12 4:37 PM, Sunil Mushran wrote:
> In 1.4, the local allocator window is small. 8MB. Meaning the node
> has to hit the global bitmap after every 8MB. In later releases, the
> window is much larger.
> 

At the moment, the OCFS2 project page
(http://oss.oracle.com/projects/ocfs2/) indicates:

"[OCFS2 + RHEL] support is limited to OCFS2 versions 1.2 and 1.4. Later
versions are not available for RHEL."

(I haven't read the related My Oracle Support doc ID yet, but the
statement above sounds pretty unambiguous.)

-------

May I ask the RHEL5 sysadmins on ocfs2-users if/how you're running OCFS2
1.6 on your systems? (Git tracking / building your own?)


> Second, a single node is not a good baseline. A better baseline is
> multiple nodes writing concurrently to the block device. Not fs.
> Use dd. Set different write offsets. This should help figure out how
> the shared device works with multiple nodes.
> 

I'm working on a new baseline (both nodes against the block device, with
appropriate write offsets) now. I will compare that to the OCFS2 tests.

Best regards -

-- 
Erik Schwartz <schwartz.erik.c at gmail.com> | GPG key 14F1139B