[
https://issues.apache.org/jira/browse/HDFS-11187?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Wei-Chiu Chuang updated HDFS-11187:
-----------------------------------
Attachment: HDFS-11187.002.patch
Posted a new patch.
In this patch, if a BlockSender sends a Finalized replica, it checks to see if
the last partial checksum is in memory. If yes, it uses the in-memory checksum.
Otherwise, it loads the checksum from disk. This is done outside the dataset
lock so should eliminate most of performance issue associated with a slow/busy
disk.
The last partial checksum of a finalized replica is stored in-memory when a
replica is converted from finalized/rbw/rur, since the checksum is already
in-memory, and does not require additional disk access. There are other cases
where a Finalized replica object is created, for example, when a DataNode
starts. In these cases it would require additional disk access to get checksum,
and the replica may not even be accessed at all, so it seems a "lazy" approach
where checksum is loaded when replica is accessed makes more sense.
> Optimize disk access for last partial chunk checksum of Finalized replica
> -------------------------------------------------------------------------
>
> Key: HDFS-11187
> URL: https://issues.apache.org/jira/browse/HDFS-11187
> Project: Hadoop HDFS
> Issue Type: Improvement
> Components: datanode
> Reporter: Wei-Chiu Chuang
> Assignee: Wei-Chiu Chuang
> Attachments: HDFS-11187.001.patch, HDFS-11187.002.patch
>
>
> The patch at HDFS-11160 ensures BlockSender reads the correct version of
> metafile when there are concurrent writers.
> However, the implementation is not optimal, because it must always read the
> last partial chunk checksum from disk while holding FsDatasetImpl lock for
> every reader. It is possible to optimize this by keeping an up-to-date
> version of last partial checksum in-memory and reduce disk access.
> I am separating the optimization into a new jira, because maintaining the
> state of in-memory checksum requires a lot more work.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
