[
https://issues.apache.org/jira/browse/HDFS-14402?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16839033#comment-16839033
]
Rakesh R commented on HDFS-14402:
---------------------------------
Thanks [~PhiloHe] for taking this jira ahead. Please go through below review
comments on the patch:
# Checksum verification uses {{blockChannel}} and reads content at the replica
device throughput. Again, reading block content make the replica device busy
and will affect other client read/write operation, assume replica device is a
HDD. Since the data is written to NVMe, you can read it back from NVMe. Also,
this will act as a verification checkpoint to ensure there is no data loss
after writing the block to NVMe device, right?
{code}
verifyChecksum(length, metaIn, blockChannel, blockFileName);
{code}
# Can we make checksum compute on/off based on some configuration and this can
improve performance and not make the device busy. We can introduce this tuning
on/off parameter applicable to all MappableBlockLoaders.
# Please change {{protected int fillBuffer}} to {{private}} visibility.
> Improve the implementation for HDFS cache on SCM
> ------------------------------------------------
>
> Key: HDFS-14402
> URL: https://issues.apache.org/jira/browse/HDFS-14402
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Components: caching, datanode
> Reporter: Feilong He
> Assignee: Feilong He
> Priority: Major
> Attachments: HDFS-14402.000.patch, HFDS-14402.001.patch
>
>
> We will consider to use transferTo API to improve SCM's cach performace.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
