You are viewing a plain text version of this content. The canonical link for it is here.
Posted to common-dev@hadoop.apache.org by "dhruba borthakur (JIRA)" <ji...@apache.org> on 2007/03/14 20:15:09 UTC
[jira] Commented: (HADOOP-971) Improve name node performance by
adding a hostname to datanodes map
[ https://issues.apache.org/jira/browse/HADOOP-971?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12480877 ]
dhruba borthakur commented on HADOOP-971:
-----------------------------------------
I saw getDatanodeByHost() consumes appreciable amount of CPU on a 1800 node cluster.
> Improve name node performance by adding a hostname to datanodes map
> -------------------------------------------------------------------
>
> Key: HADOOP-971
> URL: https://issues.apache.org/jira/browse/HADOOP-971
> Project: Hadoop
> Issue Type: Improvement
> Components: dfs
> Affects Versions: 0.11.0
> Reporter: Hairong Kuang
> Assigned To: Hairong Kuang
>
> A name node currently maintains a datanode map which maps a storage id to a dataNodeDescriptor. When the name node needs to get a datanodeDescriptor by its name (hostname:port#) or by its host name, it has to iterate through the data node set, which is very inefficent. I'd like to add an additional map that maps a host name to an array list of DatanodeDescriptors that contains all the data nodes running on the host. Since most of time a node runs only one data node, the intial size of the array list is set to be 1.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.