[
https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12764594#action_12764594
]
Todd Lipcon commented on HDFS-347:
----------------------------------
Spent some more time on the implementation this weekend. Here's a benchmark
including checksums:
t...@todd-laptop:~/git/hadoop-common/build/hadoop-core-0.21.0-dev$ time
./bin/hadoop fs -Ddfs.client.use.unix.sockets=false -cat bigfile bigfile
bigfile > /dev/null
real 0m13.502s
user 0m9.561s
sys 0m2.904s
t...@todd-laptop:~/git/hadoop-common/build/hadoop-core-0.21.0-dev$ time
./bin/hadoop fs -Ddfs.client.use.unix.sockets=true -cat bigfile bigfile bigfile
> /dev/null
real 0m9.644s
user 0m8.321s
sys 0m1.012s
bigfile is a 700MB file that I put on my local pseudo-distributed cluster.
For comparison, here's just catting the same file:
t...@todd-laptop:~/git/hadoop-common/build/hadoop-core-0.21.0-dev$ time
./bin/hadoop fs -cat file:///var/www/ubuntu-8.10-desktop-amd64.iso
file:///var/www/ubuntu-8.10-desktop-amd64.iso
file:///var/www/ubuntu-8.10-desktop-amd64.iso > /dev/null
real 0m2.914s
user 0m1.760s
sys 0m1.068s
So, the result is about a 30% speedup over the current implementation, but
still 3x overhead compared to local filesystem. Profiling shows that most of
this is in the copying out of the direct FileChannels - I think a little bit of
smart buffering there to create larger reads over the native-code boundary will
get us closer to 2x overhead.
Will clean up the patch tomorrow and upload (though it still needs plenty of
work to be commitable)
> DFS read performance suboptimal when client co-located on nodes with data
> -------------------------------------------------------------------------
>
> Key: HDFS-347
> URL: https://issues.apache.org/jira/browse/HDFS-347
> Project: Hadoop HDFS
> Issue Type: Improvement
> Reporter: George Porter
> Attachments: HADOOP-4801.1.patch, HADOOP-4801.2.patch,
> HADOOP-4801.3.patch, local-reads-doc
>
>
> One of the major strategies Hadoop uses to get scalable data processing is to
> move the code to the data. However, putting the DFS client on the same
> physical node as the data blocks it acts on doesn't improve read performance
> as much as expected.
> After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem
> is due to the HDFS streaming protocol causing many more read I/O operations
> (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB
> disk block from the DataNode process (running in a separate JVM) running on
> the same machine. The DataNode will satisfy the single disk block request by
> sending data back to the HDFS client in 64-KB chunks. In BlockSender.java,
> this is done in the sendChunk() method, relying on Java's transferTo()
> method. Depending on the host O/S and JVM implementation, transferTo() is
> implemented as either a sendfilev() syscall or a pair of mmap() and write().
> In either case, each chunk is read from the disk by issuing a separate I/O
> operation for each chunk. The result is that the single request for a 64-MB
> block ends up hitting the disk as over a thousand smaller requests for 64-KB
> each.
> Since the DFSClient runs in a different JVM and process than the DataNode,
> shuttling data from the disk to the DFSClient also results in context
> switches each time network packets get sent (in this case, the 64-kb chunk
> turns into a large number of 1500 byte packet send operations). Thus we see
> a large number of context switches for each block send operation.
> I'd like to get some feedback on the best way to address this, but I think
> providing a mechanism for a DFSClient to directly open data blocks that
> happen to be on the same machine. It could do this by examining the set of
> LocatedBlocks returned by the NameNode, marking those that should be resident
> on the local host. Since the DataNode and DFSClient (probably) share the
> same hadoop configuration, the DFSClient should be able to find the files
> holding the block data, and it could directly open them and send data back to
> the client. This would avoid the context switches imposed by the network
> layer, and would allow for much larger read buffers than 64KB, which should
> reduce the number of iops imposed by each read block operation.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
