hadoop-common-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Tsz Wo (Nicholas), SZE (JIRA)" <j...@apache.org>
Subject [jira] Updated: (HADOOP-2423) The codes in FSDirectory.mkdirs(...) is inefficient.
Date Tue, 04 Mar 2008 18:11:41 GMT

     [ https://issues.apache.org/jira/browse/HADOOP-2423?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Tsz Wo (Nicholas), SZE updated HADOOP-2423:
-------------------------------------------

    Attachment: 2423_20080304.patch

2423_20080304.patch:

The ops per sec in NNThroughputBenchmark create increases from 82.6 to 107.0 in my machine.

> The codes in FSDirectory.mkdirs(...) is inefficient.
> ----------------------------------------------------
>
>                 Key: HADOOP-2423
>                 URL: https://issues.apache.org/jira/browse/HADOOP-2423
>             Project: Hadoop Core
>          Issue Type: Improvement
>          Components: dfs
>    Affects Versions: 0.15.1
>            Reporter: Tsz Wo (Nicholas), SZE
>            Assignee: Tsz Wo (Nicholas), SZE
>         Attachments: 2423_20080130.patch, 2423_20080303.patch, 2423_20080304.patch
>
>
> FSDirectory.mkdirs(...) creates List<String> v to store all dirs.  e.g.
> {code}
> //Suppose 
> src = "/foo/bar/bas/"
> //Then,
> v = {"/", "/foo", "/foo/bar", "/foo/bar/bas"}
> {code}
> For each directory string *cur* in v, no matter *cur* already exists or not, it will
try to do a unprotectedMkdir(cur, ...).  Then, *cur* is parsed to byte[][] in INodeDirectory.addNode
(...).
> We don't need to do the parsing for each string in v.  Instead, byte[][] should be stored.
 Also, the loop should not continue once it finds an existing subdirectory.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message