[jira] Commented: (HADOOP-3856) Asynchronous IO Handling in Hadoop and HDFS

["Meng Mao (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/HADOOP-3856?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12637647#action_12637647 ] 

Meng Mao commented on HADOOP-3856:
----------------------------------

Any news on this front?

Our team has a vested interest in an alternate solution to the current DFS data write implementation. Couching my argument in a very naive perspective:
Our code worked just fine in 0.15, with no apparent inability to scale, even on very large test data sets. We've been stuck trying to get those same tests to work in 0.18 for ages, and at least are now aware of the design changes that are causing failures. What it boils down to for us, is that if Hadoop sticks with the 2 threads per write operation model, then our code will simply outstrip machine resources when we try to scale. So our course of action would have to be to review each mapreduce operation in our pipeline and consider whether it might incur a high number of threads.

Conversely, it'd be much easier for us to consume a version of Hadoop where #threads doesn't grow with #write operations. Or at least not in a linear fashion. We'd be happy to help try out any patches toward this, including the those posted in this issue. It's a matter of whether they're ready to go, I guess.

> Asynchronous IO Handling in Hadoop and HDFS
> -------------------------------------------
>
>                 Key: HADOOP-3856
>                 URL: https://issues.apache.org/jira/browse/HADOOP-3856
>             Project: Hadoop Core
>          Issue Type: New Feature
>          Components: dfs, io
>            Reporter: Raghu Angadi
>            Assignee: Raghu Angadi
>         Attachments: GrizzlyEchoServer.patch, MinaEchoServer.patch
>
>
> I think Hadoop needs utilities or framework to make it simpler to deal with generic asynchronous IO in  Hadoop.
> Example use case :
> Its been a long standing problem that DataNode takes too many threads for data transfers. Each write operation takes up 2 threads at each of the datanodes and each read operation takes one irrespective of how much activity is on the sockets. The kinds of load that HDFS serves has been expanding quite fast and HDFS should handle these varied loads better. If there is a framework for non-blocking IO, read and write pipeline state machines could be implemented with async events on a fixed number of threads. 
> A generic utility is better since it could be used in other places like DFSClient. DFSClient currently creates 2 extra threads for each file it has open for writing.
> Initially I started writing a primitive "selector", then tried to see if such facility already exists. [Apache MINA|http://mina.apache.org] seemed to do exactly this. My impression after looking the the interface and examples is that it does not give kind control we might prefer or need.  First use case I was thinking of implementing using MINA was to replace "response handlers" in DataNode. The response handlers are simpler since they don't involve disk I/O. I [asked on MINA user list|http://www.nabble.com/Async-events-with-existing-NIO-sockets.-td18640767.html], but looks like it can not be done, I think mainly because the sockets are already created.
> Essentially what I have in mind is similar to MINA, except that read and write of the sockets is done by the event handlers. The lowest layer essentially invokes selectors, invokes event handlers on single or on multiple threads. Each event handler is is expected to do some non-blocking work. We would of course have utility handler implementations that do  read, write, accept etc, that are useful for simple processing.
> Sam Pullara mentioned that [xSockets|http://xsocket.sourceforge.net/] is more flexible. It is under GPL.
> Are there other such implementations we should look at?

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.