[ https://issues.apache.org/jira/browse/HDFS-16009?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17340575#comment-17340575 ]
JiangHua Zhu edited comment on HDFS-16009 at 5/7/21, 5:38 AM: -------------------------------------------------------------- Hi [~LeonG]. I noticed that [HDFS-15547|https://issues.apache.org/jira/browse/HDFS-15547] and [HDFS-15683|https://issues.apache.org/jira/browse/HDFS-15683] has solved a similar problem, which is a good feature. was (Author: jianghuazhu): Hi [~LeonG]. I noticed that [HDFS-15547|https://issues.apache.org/jira/browse/HDFS-15547] and [HDFS-15683|https://issues.apache.org/jira/browse/HDFS-15683] has solved a similar problem, which is a good feature. > HDFS tiered storage support > --------------------------- > > Key: HDFS-16009 > URL: https://issues.apache.org/jira/browse/HDFS-16009 > Project: Hadoop HDFS > Issue Type: Improvement > Reporter: JiangHua Zhu > Priority: Major > > HDFS currently stores data blocks on different data nodes. These data nodes > are random within the scope of the network topology (local rack/different > racks/node group). > When the NameNode allocates DataNodes to the Block, it is randomly selected, > that is, each DataNode is selected equally. > We should make a hierarchical strategy. For example, DataNode1, DataNode2, > and DataNode3 are the same level, and DataNode4 and DataNode5 are the same > level. Because the hardware between DataNodes may be different, in this case, > it is necessary for us to do tiered storage, so that DataNodes with poor > hardware can store up to 50%, and 40% or even lower. This is conducive to > cluster maintenance and stability. -- This message was sent by Atlassian Jira (v8.3.4#803005) --------------------------------------------------------------------- To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org