[
https://issues.apache.org/jira/browse/HDFS-6482?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14066988#comment-14066988
]
Colin Patrick McCabe commented on HDFS-6482:
--------------------------------------------
1 second for 100k blocks is pretty good.
bq. Added a configuration parameter for users to specify the number of threads
to be used in the hard link process.
Perhaps one thread per storage directory would make sense? I'm not sure if a
configuration option is useful, if this upgrade is a one time event (and the
NameNodes that would be upgraded have already been deployed.)
bq. We use these optimizations for the hard link process only when upgrading to
the block ID-based layout, because otherwise the directory structures of the
old and new layouts should be the same and we can perform fast batch hard links
over directories – see HDFS-1445.
Why not always use the native path, if it's faster? It should be trivial to
implement the "batch" symlink API via the native path. You'd just write a
"for" loop in java that made some calls down into the JNI function you already
wrote. There is a new symlink API coming up in Java7, so we'll want to stop
using the shell thing eventually anyway.
> Use block ID-based block layout on datanodes
> --------------------------------------------
>
> Key: HDFS-6482
> URL: https://issues.apache.org/jira/browse/HDFS-6482
> Project: Hadoop HDFS
> Issue Type: Improvement
> Components: datanode
> Affects Versions: 2.5.0
> Reporter: James Thomas
> Assignee: James Thomas
> Attachments: 6482-design.doc, HDFS-6482.1.patch, HDFS-6482.2.patch,
> HDFS-6482.3.patch, HDFS-6482.4.patch, HDFS-6482.5.patch, HDFS-6482.6.patch,
> HDFS-6482.7.patch, HDFS-6482.8.patch, HDFS-6482.patch
>
>
> Right now blocks are placed into directories that are split into many
> subdirectories when capacity is reached. Instead we can use a block's ID to
> determine the path it should go in. This eliminates the need for the LDir
> data structure that facilitates the splitting of directories when they reach
> capacity as well as fields in ReplicaInfo that keep track of a replica's
> location.
> An extension of the work in HDFS-3290.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
