hadoop-hdfs-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Todd Lipcon (Commented) (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HDFS-2834) ByteBuffer-based read API for DFSInputStream
Date Tue, 06 Mar 2012 07:22:58 GMT

    [ https://issues.apache.org/jira/browse/HDFS-2834?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13223052#comment-13223052
] 

Todd Lipcon commented on HDFS-2834:
-----------------------------------

Hi Dhruba. Looks like your implementation provides for "zero copy" reads, since you can return
mmapped memory directly. This is a neat trick, but a little different from this JIRA, where
it still copies (just once) into a user-provided buffer. Also, looks like your implementation
doesn't do any checksumming for this API, right? Other concerns with the mmap approach are:
- it doesn't actually unmap when it goes "out of scope" as your comment indicates -- you need
to wait on an actual GC to call the finalizers, which can cause the process to run out of
address space on a 32-bit JVM if there isn't actual pressure on the Java heap
- the mmap() call will cause a TLB shootdown across all the threads - I'd be surprised if
the API is actually faster for the case of smaller multi-threaded reads. See section 5.1 of
this paper for more info: http://www.scribd.com/doc/59150636/C4-Continuously-Concurrent-Compacting-Collector
                
> ByteBuffer-based read API for DFSInputStream
> --------------------------------------------
>
>                 Key: HDFS-2834
>                 URL: https://issues.apache.org/jira/browse/HDFS-2834
>             Project: Hadoop HDFS
>          Issue Type: Improvement
>            Reporter: Henry Robinson
>            Assignee: Henry Robinson
>         Attachments: HDFS-2834-no-common.patch, HDFS-2834.3.patch, HDFS-2834.4.patch,
HDFS-2834.5.patch, HDFS-2834.6.patch, HDFS-2834.patch, HDFS-2834.patch, hdfs-2834-libhdfs-benchmark.png
>
>
> The {{DFSInputStream}} read-path always copies bytes into a JVM-allocated {{byte[]}}.
Although for many clients this is desired behaviour, in certain situations, such as native-reads
through libhdfs, this imposes an extra copy penalty since the {{byte[]}} needs to be copied
out again into a natively readable memory area. 
> For these cases, it would be preferable to allow the client to supply its own buffer,
wrapped in a {{ByteBuffer}}, to avoid that final copy overhead. 

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira

        

Mime
View raw message