hadoop-hdfs-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Suresh Srinivas (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HDFS-1110) Namenode heap optimization - reuse objects for commonly used file names
Date Thu, 13 May 2010 06:15:05 GMT

    [ https://issues.apache.org/jira/browse/HDFS-1110?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12867064#action_12867064

Suresh Srinivas commented on HDFS-1110:

bq. Maybe you can have only one dictionary where you store the count-of-occurances as well.
i.e move the count from the transient map to the dictionary...
I prefer having two different hashmaps for the following reasons:
# If a single hashmap is used, it grows to the size of all the names, not just the names stored
in the dictionary. Either we have shrink the map post purging entries (by making another copy)
or use heap more than necessary to retain a map much larger than what it should be.
# In dictionary, I do not intend to track the number of times the name is used. This information
is unnecessary. All we care about is whether it used more than certain threshold and not the
exact number of times a name is used.

> Namenode heap optimization - reuse objects for commonly used file names
> -----------------------------------------------------------------------
>                 Key: HDFS-1110
>                 URL: https://issues.apache.org/jira/browse/HDFS-1110
>             Project: Hadoop HDFS
>          Issue Type: Improvement
>            Reporter: Suresh Srinivas
>            Assignee: Suresh Srinivas
>             Fix For: 0.22.0
>         Attachments: hdfs-1110.2.patch, hdfs-1110.patch
> There are a lot of common file names used in HDFS, mainly created by mapreduce, such
as file names starting with "part". Reusing byte[] corresponding to these recurring file names
will save significant heap space used for storing the file names in millions of INodeFile

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

View raw message