| From: |
| Mingming Cao <cmm@us.ibm.com> |
| To: |
| ext2-devel <ext2-devel@lists.sourceforge.net>,
Andrew Morton <akpm@osdl.org>,
"Stephen C. Tweedie" <sct@redhat.com>,
linux-kernel <linux-kernel@vger.kernel.org>,
linux-fsdevel@vger.kernel.org |
| Subject: |
| [RFC] [PATCH 0/4]Multiple block allocation and delayed allocation
for ext3 |
| Date: |
| Sun, 17 Jul 2005 10:40:02 -0700 |
| Cc: |
| Badari Pulavarty <pbadari@us.ibm.com>, suparna@in.ibm.com,
tytso@mit.edu, alex@clusterfs.com, adilger@clusterfs.com |
Hi All,
Here are the updated patches to support multiple block allocation and
delayed allocation for ext3 done by me, Badari and Suparna.
[PATCH 1/4] -- multiple block allocation for current ext3.
(ext3_get_blocks()).
[PATCH 2/4] -- adding delayed allocation for writeback mode
[PATCH 3/4] -- generic support for cluster pages together in
mapge_writepages() to make use of getblocks()
[PATCH 4/4] -- support multiple block allocation for ext3 writeback mode
through writepages().
Have done initial testing on dbench and tiobench on a 2.6.11 version of
this patch set. Dbench 8 thread throughput result is increased by 20%
with this patch set.
dbench comparison: (ext3-dm represents ext3+thispatchset)
http://www.sudhaa.com/~ram/ols2005presentation/dbench.jpg
tiobench comparison:
http://www.sudhaa.com/~ram/ols2005presentation/tio_seq_wr...
Todo:
- bmap() support for delayed allocation
- page reserve flag to indicate the delayed allocation
- ordered mode support for delayed allocation
- "bh" support to enable blocksize = 1k/2k filesystems
Cheers,
Mingming
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/