hadoop-hdfs-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Ray Zhang (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HDFS-7737) Implement distributed namenode
Date Wed, 11 Mar 2015 03:05:39 GMT

    [ https://issues.apache.org/jira/browse/HDFS-7737?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14356147#comment-14356147
] 

Ray Zhang commented on HDFS-7737:
---------------------------------

Have finished prototype without replication feature, will run performance test after bug fix
and share the test result. Will compare performance with HDFS Federation (static subtree partitioning).

> Implement distributed namenode
> ------------------------------
>
>                 Key: HDFS-7737
>                 URL: https://issues.apache.org/jira/browse/HDFS-7737
>             Project: Hadoop HDFS
>          Issue Type: New Feature
>            Reporter: Ray Zhang
>            Assignee: Ray Zhang
>            Priority: Minor
>         Attachments: Distributed Namenode.pdf, Dynamic Subtree Binary Partitioning Proposal.pdf,
Implement Overflow Table.pdf, Replication Of Name Nodes.pdf
>
>
> Try to add the following functions on HDFS:
> 1) Manually move metadata across different name nodes, without losing the original data
locality
> 2) Automatically balance metadata across different name nodes with a novel subtree partition
algorithm
> 3) Able to set a replication factor number for each name node, and each replicated name
node can balance live read/write traffic
> 4) All the functionalities above are highly configurable, and the administrator can decide
if he wants to use distributed name node setup or just HDFS federation



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message