[
https://issues.apache.org/jira/browse/RATIS-1176?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17238163#comment-17238163
]
runzhiwang edited comment on RATIS-1176 at 11/24/20, 11:48 PM:
---------------------------------------------------------------
bq. I suspect it is better than transferTo(..) since it is using
transferToArbitraryChannel(..).
I agree. I think ratis streaming is faster than transferToArbitraryChannel,
slower than transferToTrustedChannel and transferToDirectly. Because when
primary/peer receive data in ratis streaming, it need DirectByteBuffer in netty
to save data, as the following image shows, so primary/peer need to read data
by flow: socket -> kernel space -> user space, and when primary send data to
peer, the flow is: user space -> kernel space -> socket, when primary/peer save
data to disk, the flow is: user space -> kernel space -> disk, so I think ratis
streaming is slower than transferToTrustedChannel which did not need copy
between user space and kernel space in primary and peer.
!image-2020-11-25-07-40-50-383.png!
bq. We may pass MapByteBuffer to our writeAsync(..) method
I agree.
was (Author: yjxxtd):
bq. I suspect it is better than transferTo(..) since it is using
transferToArbitraryChannel(..).
I agree. I think ratis streaming is faster than transferToArbitraryChannel, not
sure whether faster than transferToTrustedChannel, and slower than
transferToDirectly. Because when primary/peer receive data in ratis streaming,
it need DirectByteBuffer in netty to save data, as the following image shows,
so primary/peer need to read data by flow: socket -> kernel space -> user
space, and when primary send data to peer, the flow is: user space -> kernel
space -> socket, when primary/peer save data to disk, the flow is: user space
-> kernel space -> disk, so I think ratis streaming is slower than
transferToDirectly which did not need copy between user space and kernel space
in primary and peer.
!image-2020-11-25-07-40-50-383.png!
bq. We may pass MapByteBuffer to our writeAsync(..) method
I agree.
> Benchmark various ways to stream data
> -------------------------------------
>
> Key: RATIS-1176
> URL: https://issues.apache.org/jira/browse/RATIS-1176
> Project: Ratis
> Issue Type: Sub-task
> Components: client, Streaming
> Reporter: Tsz-wo Sze
> Priority: Major
> Attachments: image-2020-11-25-07-40-50-383.png
>
>
> In RATIS-1175, we provided a WritableByteChannel view of DataStreamOutput in
> order to support FileChannel.transferTo. However, [~runzhiwang] pointed out
> that sun.nio.ch.FileChannelImpl.transferTo has three submethods
> - transferToDirectly (fastest)
> - transferToTrustedChannel
> - transferToArbitraryChannel (slowest, requires buffer copying)
> Unfortunately, our current implementation only able to use
> transferToArbitraryChannel.
> There are several ideas below to improve the performance. We should
> benchmark them.
> # Improve the current implementation of WritableByteChannel so that it may be
> able to use a faster transferTo method.
> # Use
> [FileChannel.map(..)|https://docs.oracle.com/javase/8/docs/api/java/nio/channels/FileChannel.html#map-java.nio.channels.FileChannel.MapMode-long-long-]
> and pass MappedByteBuffer to our DataStreamOutput.writeAsync method.
> # Add a new API
> {code}
> //DataStreamOutput
> CompletableFuture<DataStreamReply> writeAsync(File);
> {code}
> Internally, use Netty DefaultFileRegion for zero-copy file transfer:
> https://github.com/netty/netty/blob/4.1/example/src/main/java/io/netty/example/file/FileServerHandler.java#L53
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
