[jira] [Commented] (IGNITE-4093) Provide implementation of java.util.concurrent.CompletionService

2016-10-20 Thread Valentin Kulichenko (JIRA)

[ 
https://issues.apache.org/jira/browse/IGNITE-4093?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15592484#comment-15592484
 ] 

Valentin Kulichenko commented on IGNITE-4093:
-

Got it. This can be an option, but I disagree that implementation with a local 
queue is broken. It's a different functionality and we can have both.

> Provide implementation of java.util.concurrent.CompletionService
> 
>
> Key: IGNITE-4093
> URL: https://issues.apache.org/jira/browse/IGNITE-4093
> Project: Ignite
>  Issue Type: Improvement
>  Components: general
>Affects Versions: 1.7
>Reporter: Valentin Kulichenko
>
> Ignite provides the implementation of `ExecutorService`, but not 
> `CompletionService`. There is `ExecutorCompletionService` in JDK, but it's 
> implementation is weird and looks like there is no way to use it in 
> distributed environment.
> I suggest to create our own implementation of `CompletionService` that will 
> execute closures in distributed fashion.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)


[jira] [Commented] (IGNITE-4093) Provide implementation of java.util.concurrent.CompletionService

2016-10-19 Thread Nadeem Mohammad (JIRA)

[ 
https://issues.apache.org/jira/browse/IGNITE-4093?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15589454#comment-15589454
 ] 

Nadeem Mohammad commented on IGNITE-4093:
-

Infinispan's org.infinispan.distexec.DistributedExecutionCompletionService 
internally uses LinkedBlockingQueue, however this implementation is broken, the 
moment the node goes down (In which the queue resides), all the content of the 
nodes would be gone... further this queue is not shared with all the other 
nodes in the cluster..

So if at all a decision is made to us Blocking queue in Ignite, Then it has to 
be a Distributed Blocking Queue, otherwise the implementation would be broken 
again. hope it helps


> Provide implementation of java.util.concurrent.CompletionService
> 
>
> Key: IGNITE-4093
> URL: https://issues.apache.org/jira/browse/IGNITE-4093
> Project: Ignite
>  Issue Type: Improvement
>  Components: general
>Affects Versions: 1.7
>Reporter: Valentin Kulichenko
>
> Ignite provides the implementation of `ExecutorService`, but not 
> `CompletionService`. There is `ExecutorCompletionService` in JDK, but it's 
> implementation is weird and looks like there is no way to use it in 
> distributed environment.
> I suggest to create our own implementation of `CompletionService` that will 
> execute closures in distributed fashion.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)


[jira] [Commented] (IGNITE-4093) Provide implementation of java.util.concurrent.CompletionService

2016-10-19 Thread Valentin Kulichenko (JIRA)

[ 
https://issues.apache.org/jira/browse/IGNITE-4093?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15589381#comment-15589381
 ] 

Valentin Kulichenko commented on IGNITE-4093:
-

[~mnadeem], can you please clarify this and provide more technical details? I 
don't see any reason to have a distributed queue for the futures here.

> Provide implementation of java.util.concurrent.CompletionService
> 
>
> Key: IGNITE-4093
> URL: https://issues.apache.org/jira/browse/IGNITE-4093
> Project: Ignite
>  Issue Type: Improvement
>  Components: general
>Affects Versions: 1.7
>Reporter: Valentin Kulichenko
>
> Ignite provides the implementation of `ExecutorService`, but not 
> `CompletionService`. There is `ExecutorCompletionService` in JDK, but it's 
> implementation is weird and looks like there is no way to use it in 
> distributed environment.
> I suggest to create our own implementation of `CompletionService` that will 
> execute closures in distributed fashion.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)


[jira] [Commented] (IGNITE-4093) Provide implementation of java.util.concurrent.CompletionService

2016-10-18 Thread Nadeem Mohammad (JIRA)

[ 
https://issues.apache.org/jira/browse/IGNITE-4093?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15587862#comment-15587862
 ] 

Nadeem Mohammad commented on IGNITE-4093:
-

We have to make sure, Distributed Blocking Queue is used in the 
implementation... otherwise implementation would be broken again... This is the 
case with Infinispan... Currently no one provide one (Distributed blocking 
queue), Hazelcast provides one but that is not truly distributed.

> Provide implementation of java.util.concurrent.CompletionService
> 
>
> Key: IGNITE-4093
> URL: https://issues.apache.org/jira/browse/IGNITE-4093
> Project: Ignite
>  Issue Type: Improvement
>  Components: general
>Affects Versions: 1.7
>Reporter: Valentin Kulichenko
>
> Ignite provides the implementation of `ExecutorService`, but not 
> `CompletionService`. There is `ExecutorCompletionService` in JDK, but it's 
> implementation is weird and looks like there is no way to use it in 
> distributed environment.
> I suggest to create our own implementation of `CompletionService` that will 
> execute closures in distributed fashion.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)