[DISCUSSION] jlibaio proposal

classic Classic list List threaded Threaded
6 messages Options
Reply | Threaded
Open this post in threaded view
|

[DISCUSSION] jlibaio proposal

Clebert Suconic
I have this sub-project within ActiveMQ artemis, that is becoming more
independent and could be reused among other projects.


It's a thin layer in top of Linux libaio.

You have method to submit write, submit read, allocate byte buffers,
and they are going straight to DMA on Linux Kernel.

Right now I'm only using it on Linux, but it would be extended for
Posix Libaio on other platforms:


And since this is a nice and independent API, this could be reused by
any other project dealing with IO and wanting to go a bit faster on
the file system. We have been able to go 20% with a previous version
of this library, and it could be a bit better with the newer version.


Right now I'm going to drop this as a project on gibhub and reuse it
on artemis, but before I do that, is there any interest on a native
project on commons? (notice I looked at tomcat commons and I don't see
a place there)

---------------------------------------------------------------------
To unsubscribe, e-mail: [hidden email]
For additional commands, e-mail: [hidden email]

Reply | Threaded
Open this post in threaded view
|

Re: [DISCUSSION] jlibaio proposal

Clebert Suconic
forgot the link:

https://github.com/clebertsuconic/activemq-artemis/tree/master/artemis-native

On Mon, Jul 13, 2015 at 4:10 PM, Clebert Suconic
<[hidden email]> wrote:

> I have this sub-project within ActiveMQ artemis, that is becoming more
> independent and could be reused among other projects.
>
>
> It's a thin layer in top of Linux libaio.
>
> You have method to submit write, submit read, allocate byte buffers,
> and they are going straight to DMA on Linux Kernel.
>
> Right now I'm only using it on Linux, but it would be extended for
> Posix Libaio on other platforms:
>
>
> And since this is a nice and independent API, this could be reused by
> any other project dealing with IO and wanting to go a bit faster on
> the file system. We have been able to go 20% with a previous version
> of this library, and it could be a bit better with the newer version.
>
>
> Right now I'm going to drop this as a project on gibhub and reuse it
> on artemis, but before I do that, is there any interest on a native
> project on commons? (notice I looked at tomcat commons and I don't see
> a place there)



--
Clebert Suconic
http://community.jboss.org/people/clebert.suconic@...
http://clebertsuconic.blogspot.com

---------------------------------------------------------------------
To unsubscribe, e-mail: [hidden email]
For additional commands, e-mail: [hidden email]

Reply | Threaded
Open this post in threaded view
|

Re: [DISCUSSION] jlibaio proposal

Benedikt Ritter-4
Hello Clebert,



2015-07-13 22:11 GMT+02:00 Clebert Suconic <[hidden email]>:

> forgot the link:
>
>
> https://github.com/clebertsuconic/activemq-artemis/tree/master/artemis-native
>
> On Mon, Jul 13, 2015 at 4:10 PM, Clebert Suconic
> <[hidden email]> wrote:
> > I have this sub-project within ActiveMQ artemis, that is becoming more
> > independent and could be reused among other projects.
> >
> >
> > It's a thin layer in top of Linux libaio.
> >
> > You have method to submit write, submit read, allocate byte buffers,
> > and they are going straight to DMA on Linux Kernel.
> >
> > Right now I'm only using it on Linux, but it would be extended for
> > Posix Libaio on other platforms:
> >
> >
> > And since this is a nice and independent API, this could be reused by
> > any other project dealing with IO and wanting to go a bit faster on
> > the file system. We have been able to go 20% with a previous version
> > of this library, and it could be a bit better with the newer version.
> >
> >
> > Right now I'm going to drop this as a project on gibhub and reuse it
> > on artemis, but before I do that, is there any interest on a native
> > project on commons? (notice I looked at tomcat commons and I don't see
> > a place there)
>

Apache Commons is open to all ASF committers and we invite anybody to share
code here [1]. So generally speaking this is something that would fit into
commons. Are you the only one working on this right now? It is better to
have at lest three or four people who are working on a component.

Best regards,
Benedikt

[1]
http://mail-archives.apache.org/mod_mbox/community-dev/201412.mbox/%3CCAB917RL7gHqCmgnes8g3k0aCHMbDwnKyxN%3DBuxhtbVgSEyO0qw%40mail.gmail.com%3E


>
>
>
> --
> Clebert Suconic
> http://community.jboss.org/people/clebert.suconic@...
> http://clebertsuconic.blogspot.com
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: [hidden email]
> For additional commands, e-mail: [hidden email]
>
>


--
http://people.apache.org/~britter/
http://www.systemoutprintln.de/
http://twitter.com/BenediktRitter
http://github.com/britter
Reply | Threaded
Open this post in threaded view
|

Re: [DISCUSSION] jlibaio proposal

jochen-2
On Wed, Jul 15, 2015 at 8:43 AM, Benedikt Ritter <[hidden email]> wrote:

> Apache Commons is open to all ASF committers and we invite anybody to share
> code here [1]. So generally speaking this is something that would fit into
> commons. Are you the only one working on this right now? It is better to
> have at lest three or four people who are working on a component.

That being said, and fully aware of an existing precedence (Daemon),
but nevertheless:

Are you aware, Benedikt, that we are talking about a project, which is
written in C, C++, or something similar? Know How in that area is rare
on Commons. So, it is likely, that this subproject wouldn't fit in
very well and tend to be isolated. (Even some of us Java cracks,
including me, will find dev@commons a burden from time to time. What
do you expect a C programmer to think?)

So, although we *are* open, I'd recommend another parent project.
Like, for example, Artemis, which I'd consider to be a natural choice,
if this comes from within Artemis.

Jochen





--
Any world that can produce the Taj Mahal, William Shakespeare,
and Stripe toothpaste can't be all bad. (C.R. MacNamara, One Two Three)

---------------------------------------------------------------------
To unsubscribe, e-mail: [hidden email]
For additional commands, e-mail: [hidden email]

Reply | Threaded
Open this post in threaded view
|

Re: [DISCUSSION] jlibaio proposal

James Carman
On Wed, Jul 15, 2015 at 2:55 AM Jochen Wiedmann <[hidden email]>
wrote:

> On Wed, Jul 15, 2015 at 8:43 AM, Benedikt Ritter <[hidden email]>
> wrote:
>
> > Apache Commons is open to all ASF committers and we invite anybody to
> share
> > code here [1]. So generally speaking this is something that would fit
> into
> > commons. Are you the only one working on this right now? It is better to
> > have at lest three or four people who are working on a component.
>
> That being said, and fully aware of an existing precedence (Daemon),
> but nevertheless:
>
> Are you aware, Benedikt, that we are talking about a project, which is
> written in C, C++, or something similar? Know How in that area is rare
> on Commons. So, it is likely, that this subproject wouldn't fit in
> very well and tend to be isolated. (Even some of us Java cracks,
> including me, will find dev@commons a burden from time to time. What
> do you expect a C programmer to think?)
>
> So, although we *are* open, I'd recommend another parent project.
> Like, for example, Artemis, which I'd consider to be a natural choice,
> if this comes from within Artemis.
>
>
I would agree that I don't think it belongs in Commons.
Reply | Threaded
Open this post in threaded view
|

Re: [DISCUSSION] jlibaio proposal

sebb-2-2
On 15 July 2015 at 21:36, James Carman <[hidden email]> wrote:

> On Wed, Jul 15, 2015 at 2:55 AM Jochen Wiedmann <[hidden email]>
> wrote:
>
>> On Wed, Jul 15, 2015 at 8:43 AM, Benedikt Ritter <[hidden email]>
>> wrote:
>>
>> > Apache Commons is open to all ASF committers and we invite anybody to
>> share
>> > code here [1]. So generally speaking this is something that would fit
>> into
>> > commons. Are you the only one working on this right now? It is better to
>> > have at lest three or four people who are working on a component.
>>
>> That being said, and fully aware of an existing precedence (Daemon),
>> but nevertheless:
>>
>> Are you aware, Benedikt, that we are talking about a project, which is
>> written in C, C++, or something similar? Know How in that area is rare
>> on Commons. So, it is likely, that this subproject wouldn't fit in
>> very well and tend to be isolated. (Even some of us Java cracks,
>> including me, will find dev@commons a burden from time to time. What
>> do you expect a C programmer to think?)
>>
>> So, although we *are* open, I'd recommend another parent project.
>> Like, for example, Artemis, which I'd consider to be a natural choice,
>> if this comes from within Artemis.
>>
>>
> I would agree that I don't think it belongs in Commons.

+1

The scope of Commons is Java only:

"Apache Commons is an Apache project focused on all aspects of
reusable Java components. "

See http://commons.apache.org/

---------------------------------------------------------------------
To unsubscribe, e-mail: [hidden email]
For additional commands, e-mail: [hidden email]