[ https://issues.apache.org/jira/browse/HDFS-8019?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14528262#comment-14528262 ]
Kai Zheng commented on HDFS-8019: --------------------------------- Hi Vinay, thanks for the taking and good question. I thought it's good enough to use a simple approach for phase I as you suggested. When we have time, in DataNode side, I'm wondering if we could have dedicated erasure coding buffer management, because native direct ByteBuffer is more desired, as powerful DNs may be used to perform off-line encoding and decoding, where native erasure coders can be utilized to outperform greatly. Note native coders may use direct ByteBuffers. In either way, maybe we could have an interface for the buffer pool allocation and management, so it's easy to have a new implementation and switch to it in future. > Erasure Coding: erasure coding chunk buffer allocation and management > --------------------------------------------------------------------- > > Key: HDFS-8019 > URL: https://issues.apache.org/jira/browse/HDFS-8019 > Project: Hadoop HDFS > Issue Type: Sub-task > Reporter: Kai Zheng > Assignee: Vinayakumar B > > As a task of HDFS-7344, this is to come up a chunk buffer pool allocating and > managing coding chunk buffers, either based on on-heap or off-heap. Note this > assumes some DataNodes are powerful in computing and performing EC coding > work, so better to have this dedicated buffer pool and management. -- This message was sent by Atlassian JIRA (v6.3.4#6332)