Sequentiality of .NET driver-side events

[Jason Jeong <cu...@gmail.com>]

Hi all,

The JobDriver class in reef-bridge-java has many synchronized
(JobDriver.this) keywords, resulting in event handlers of AllocatedEvaluator
and ActiveContext events and such being fired one at a time, sequentially.
This is counter-intuitive to what we want to provide in REEF; events should
be processed concurrently, having synchronization only when the user
intends to.

I would like to know if others agree on this, or if there is some other
intended behavior behind all the synchronization.
If we agree on fixing this issue, then I would like to work on this myself,
especially since I actually have a use case blocked by this one ;)

Thanks,
Jason

Re: Sequentiality of .NET driver-side events

[Jason Jeong <cu...@gmail.com>]

Just for reference, I've created REEF-1520
<http://issues.apache.org/jira/browse/REEF-1520> for this.

Jason

On Wed, Aug 10, 2016 at 8:56 PM, Jason Jeong <cu...@gmail.com> wrote:

> Hi Markus,
>
> I'll definitely keep an eye on the .NET Linux issues and raise the alarm
> in case something happens.
> Seems that one of the approaches is getting rid of the bridge altogether?
> That indeed is another way of addressing the concurrency issue mentioned on
> this thread.
>
> In the meantime, I'll create a separate JIRA for this one.
>
> Thanks,
> Jason
>
> On Wed, Aug 10, 2016 at 8:06 PM, Markus Weimer <ma...@weimo.de> wrote:
>
>> On 2016-08-10 15:09, Jason Jeong wrote:
>>
>>> I would like to know if others agree on this, or if there is some
>>> other intended behavior behind all the synchronization.
>>>
>>
>> IIRC, the main reason for this is ease of development of the .NET / Java
>> bridge.
>>
>> If we agree on fixing this issue, then I would like to work on this
>>> myself, especially since I actually have a use case blocked by this
>>> one ;)
>>>
>>
>> Thanks! Please make sure that you coordinate with the .NET on Linux
>> thread, which might touch upon the same code base sometime this late
>> summer.
>>
>> Markus
>>
>>
>

Re: Sequentiality of .NET driver-side events

[Jason Jeong <cu...@gmail.com>]

Hi Markus,

I'll definitely keep an eye on the .NET Linux issues and raise the alarm in
case something happens.
Seems that one of the approaches is getting rid of the bridge altogether?
That indeed is another way of addressing the concurrency issue mentioned on
this thread.

In the meantime, I'll create a separate JIRA for this one.

Thanks,
Jason

On Wed, Aug 10, 2016 at 8:06 PM, Markus Weimer <ma...@weimo.de> wrote:

> On 2016-08-10 15:09, Jason Jeong wrote:
>
>> I would like to know if others agree on this, or if there is some
>> other intended behavior behind all the synchronization.
>>
>
> IIRC, the main reason for this is ease of development of the .NET / Java
> bridge.
>
> If we agree on fixing this issue, then I would like to work on this
>> myself, especially since I actually have a use case blocked by this
>> one ;)
>>
>
> Thanks! Please make sure that you coordinate with the .NET on Linux
> thread, which might touch upon the same code base sometime this late
> summer.
>
> Markus
>
>

Re: Sequentiality of .NET driver-side events

[Markus Weimer <ma...@weimo.de>]

On 2016-08-10 15:09, Jason Jeong wrote:
> I would like to know if others agree on this, or if there is some
> other intended behavior behind all the synchronization.

IIRC, the main reason for this is ease of development of the .NET / Java
bridge.

> If we agree on fixing this issue, then I would like to work on this
> myself, especially since I actually have a use case blocked by this
> one ;)

Thanks! Please make sure that you coordinate with the .NET on Linux
thread, which might touch upon the same code base sometime this late summer.

Markus