[
https://issues.apache.org/jira/browse/HBASE-5074?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13213451#comment-13213451
]
Phabricator commented on HBASE-5074:
------------------------------------
dhruba has commented on the revision "[jira] [HBASE-5074] Support checksums in
HBase block cache".
INLINE COMMENTS
src/main/java/org/apache/hadoop/hbase/fs/HFileSystem.java:115 ideally, we
need two different fs. The first fs is for writing and
reading-with-hdfs-checksums. The other fs is for reading-without-hdfs.
src/main/java/org/apache/hadoop/hbase/fs/HFileSystem.java:129 done
src/main/java/org/apache/hadoop/hbase/io/hfile/AbstractHFileReader.java:49
The HFile layer is the one that is responsible for opening a file for reading.
Then the multi-threaded HFileBlockLayer uses those FSDataInputStream to pread
data from HDFS. So, I need to make the HFile layer open two file descriptors
for the same file, both for reading purposes... one which checksum and the
other without checksums
src/main/java/org/apache/hadoop/hbase/io/hfile/AbstractHFileReader.java:44
This is a protected member, so users of this class are not concerned on what
this is. If you have a better structure on how to organize this one, please do
let me know
src/main/java/org/apache/hadoop/hbase/io/hfile/ChecksumUtil.java:84 The
Checksum API returns a long. But actual implementations like CRC32, CRC32C, etc
all return an int.
Also, the Hadoop checksum implementation also uses a 4 byte value. If you
think that we should store 8 byte checksums, I can do that. But for the common
case, we will be wasting 4 bytes in the header for every checksum chunk
src/main/java/org/apache/hadoop/hbase/io/hfile/HFile.java:205 done
REVISION DETAIL
https://reviews.facebook.net/D1521
> support checksums in HBase block cache
> --------------------------------------
>
> Key: HBASE-5074
> URL: https://issues.apache.org/jira/browse/HBASE-5074
> Project: HBase
> Issue Type: Improvement
> Components: regionserver
> Reporter: dhruba borthakur
> Assignee: dhruba borthakur
> Attachments: D1521.1.patch, D1521.1.patch, D1521.2.patch,
> D1521.2.patch, D1521.3.patch, D1521.3.patch, D1521.4.patch, D1521.4.patch,
> D1521.5.patch, D1521.5.patch, D1521.6.patch, D1521.6.patch, D1521.7.patch,
> D1521.7.patch
>
>
> The current implementation of HDFS stores the data in one block file and the
> metadata(checksum) in another block file. This means that every read into the
> HBase block cache actually consumes two disk iops, one to the datafile and
> one to the checksum file. This is a major problem for scaling HBase, because
> HBase is usually bottlenecked on the number of random disk iops that the
> storage-hardware offers.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
