[
https://issues.apache.org/jira/browse/HADOOP-13600?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16160040#comment-16160040
]
Steve Loughran commented on HADOOP-13600:
-----------------------------------------
Just had a quick look at things other than the core algorithm, which will take
more time to review than I have now
* With s3guard we want to make sure that parent things are only deleted after
all the children; that may imply a serialized bit at the end.
* we're permanently fighting file size in S3AFilesystem, which is at 3K lines
today; anything which can be done to organise stuff out of the class (progress,
common thread pool setup) helps.
* And resource use and startup time of s3a is an issue on thngs like shared
hive daemons, where an s3a instance for a single user might be instantiated for
a few reads and then deleted...we shouldn't be creating heavier things until
needed.
Maybe here we could have a transfer manager wrapper which implemented on-demand
creation, reading in the config options
{code}
XferManager transfers = new XferManager(conf, TRANSFERS, TRANSFERS_MAX)
XferManager upload = new XferManager(conf, UPLOADS, UPLOADS_MAX)
// and with the .get() method doing on-demand instantiation of things including
thread pools as needed.
transfers.get().startTransfer(...)
{code}
> S3a rename() to copy files in a directory in parallel
> -----------------------------------------------------
>
> Key: HADOOP-13600
> URL: https://issues.apache.org/jira/browse/HADOOP-13600
> Project: Hadoop Common
> Issue Type: Sub-task
> Components: fs/s3
> Affects Versions: 2.7.3
> Reporter: Steve Loughran
> Assignee: Sahil Takiar
>
> Currently a directory rename does a one-by-one copy, making the request
> O(files * data). If the copy operations were launched in parallel, the
> duration of the copy may be reducable to the duration of the longest copy.
> For a directory with many files, this will be significant
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
