Re: [DISCUSS] Making submarine to different release model like Ozone

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

Re: [DISCUSS] Making submarine to different release model like Ozone

John Zhuge-2
+1

Does Submarine support Jupyter?

On Fri, Feb 1, 2019 at 8:54 AM Zhe Zhang <[hidden email]> wrote:

> +1 on the proposal and looking forward to the progress of the project!
>
> On Thu, Jan 31, 2019 at 10:51 PM Weiwei Yang <[hidden email]> wrote:
>
> > Thanks for proposing this Wangda, my +1 as well.
> > It is amazing to see the progress made in Submarine last year, the
> > community grows fast and quiet collaborative. I can see the reasons to
> get
> > it release faster in its own cycle. And at the same time, the Ozone way
> > works very well.
> >
> > —
> > Weiwei
> > On Feb 1, 2019, 10:49 AM +0800, Xun Liu <[hidden email]>, wrote:
> > > +1
> > >
> > > Hello everyone,
> > >
> > > I am Xun Liu, the head of the machine learning team at Netease Research
> > Institute. I quite agree with Wangda.
> > >
> > > Our team is very grateful for getting Submarine machine learning engine
> > from the community.
> > > We are heavy users of Submarine.
> > > Because Submarine fits into the direction of our big data team's hadoop
> > technology stack,
> > > It avoids the needs to increase the manpower investment in learning
> > other container scheduling systems.
> > > The important thing is that we can use a common YARN cluster to run
> > machine learning,
> > > which makes the utilization of server resources more efficient, and
> > reserves a lot of human and material resources in our previous years.
> > >
> > > Our team have finished the test and deployment of the Submarine and
> will
> > provide the service to our e-commerce department (http://www.kaola.com/)
> > shortly.
> > >
> > > We also plan to provides the Submarine engine in our existing YARN
> > cluster in the next six months.
> > > Because we have a lot of product departments need to use machine
> > learning services,
> > > for example:
> > > 1) Game department (http://game.163.com/) needs AI battle training,
> > > 2) News department (http://www.163.com) needs news recommendation,
> > > 3) Mailbox department (http://www.163.com) requires anti-spam and
> > illegal detection,
> > > 4) Music department (https://music.163.com/) requires music
> > recommendation,
> > > 5) Education department (http://www.youdao.com) requires voice
> > recognition,
> > > 6) Massive Open Online Courses (https://open.163.com/) requires
> > multilingual translation and so on.
> > >
> > > If Submarine can be released independently like Ozone, it will help us
> > quickly get the latest features and improvements, and it will be great
> > helpful to our team and users.
> > >
> > > Thanks hadoop Community!
> > >
> > >
> > > > 在 2019年2月1日,上午2:53,Wangda Tan <[hidden email]> 写道:
> > > >
> > > > Hi devs,
> > > >
> > > > Since we started submarine-related effort last year, we received a
> lot
> > of
> > > > feedbacks, several companies (such as Netease, China Mobile, etc.)
> are
> > > > trying to deploy Submarine to their Hadoop cluster along with big
> data
> > > > workloads. Linkedin also has big interests to contribute a Submarine
> > TonY (
> > > > https://github.com/linkedin/TonY) runtime to allow users to use the
> > same
> > > > interface.
> > > >
> > > > From what I can see, there're several issues of putting Submarine
> under
> > > > yarn-applications directory and have same release cycle with Hadoop:
> > > >
> > > > 1) We started 3.2.0 release at Sep 2018, but the release is done at
> Jan
> > > > 2019. Because of non-predictable blockers and security issues, it got
> > > > delayed a lot. We need to iterate submarine fast at this point.
> > > >
> > > > 2) We also see a lot of requirements to use Submarine on older Hadoop
> > > > releases such as 2.x. Many companies may not upgrade Hadoop to 3.x
> in a
> > > > short time, but the requirement to run deep learning is urgent to
> > them. We
> > > > should decouple Submarine from Hadoop version.
> > > >
> > > > And why we wanna to keep it within Hadoop? First, Submarine included
> > some
> > > > innovation parts such as enhancements of user experiences for YARN
> > > > services/containerization support which we can add it back to Hadoop
> > later
> > > > to address common requirements. In addition to that, we have a big
> > overlap
> > > > in the community developing and using it.
> > > >
> > > > There're several proposals we have went through during Ozone merge to
> > trunk
> > > > discussion:
> > > >
> >
> https://mail-archives.apache.org/mod_mbox/hadoop-common-dev/201803.mbox/%3CCAHfHakH6_m3YLdf5a2KQ8+w-5fbVX5aHFgS-x1VaJW8gmnzRLg@...%3E
> > > >
> > > > I propose to adopt Ozone model: which is the same master branch,
> > different
> > > > release cycle, and different release branch. It is a great example to
> > show
> > > > agile release we can do (2 Ozone releases after Oct 2018) with less
> > > > overhead to setup CI, projects, etc.
> > > >
> > > > *Links:*
> > > > - JIRA: https://issues.apache.org/jira/browse/YARN-8135
> > > > - Design doc
> > > > <
> >
> https://docs.google.com/document/d/199J4pB3blqgV9SCNvBbTqkEoQdjoyGMjESV4MktCo0k/edit
> > >
> > > > - User doc
> > > > <
> >
> https://hadoop.apache.org/docs/r3.2.0/hadoop-yarn/hadoop-yarn-applications/hadoop-yarn-submarine/Index.html
> > >
> > > > (3.2.0
> > > > release)
> > > > - Blogposts, {Submarine} : Running deep learning workloads on Apache
> > Hadoop
> > > > <
> >
> https://hortonworks.com/blog/submarine-running-deep-learning-workloads-apache-hadoop/
> > >,
> > > > (Chinese Translation: Link <https://www.jishuwen.com/d/2Vpu>)
> > > > - Talks: Strata Data Conf NY
> > > > <
> >
> https://conferences.oreilly.com/strata/strata-ny-2018/public/schedule/detail/68289
> > >
> > > >
> > > > Thoughts?
> > > >
> > > > Thanks,
> > > > Wangda Tan
> > >
> > >
> > >
> > > ---------------------------------------------------------------------
> > > To unsubscribe, e-mail: [hidden email]
> > > For additional commands, e-mail: [hidden email]
> > >
> >
>


--
John
Reply | Threaded
Open this post in threaded view
|

Re: [DISCUSS] Making submarine to different release model like Ozone

Ajay Kumar
+1, Thanks for driving this. With rise of use cases running ML along with traditional applications this will be of great help.

Thanks,
Ajay  

On 2/1/19, 10:49 AM, "Suma Shivaprasad" <[hidden email]> wrote:

    +1. Thanks for bringing this up Wangda.
   
    Makes sense to have Submarine follow its own release cadence given the good
    momentum/adoption so far. Also, making it run with older versions of Hadoop
    would drive higher adoption.
   
    Suma
   
    On Fri, Feb 1, 2019 at 9:40 AM Eric Yang <[hidden email]> wrote:
   
    > Submarine is an application built for YARN framework, but it does not have
    > strong dependency on YARN development.  For this kind of projects, it would
    > be best to enter Apache Incubator cycles to create a new community.  Apache
    > commons is the only project other than Incubator that has independent
    > release cycles.  The collection is large, and the project goal is
    > ambitious.  No one really knows which component works with each other in
    > Apache commons.  Hadoop is a much more focused project on distributed
    > computing framework and not incubation sandbox.  For alignment with Hadoop
    > goals, and we want to prevent Hadoop project to be overloaded while
    > allowing good ideas to be carried forwarded in Apache incubator.  Put on my
    > Apache Member hat, my vote is -1 to allow more independent subproject
    > release cycle in Hadoop project that does not align with Hadoop project
    > goals.
    >
    > Apache incubator process is highly recommended for Submarine:
    > https://incubator.apache.org/policy/process.html This allows Submarine to
    > develop for older version of Hadoop like Spark works with multiple versions
    > of Hadoop.
    >
    > Regards,
    > Eric
    >
    > On 1/31/19, 10:51 PM, "Weiwei Yang" <[hidden email]> wrote:
    >
    >     Thanks for proposing this Wangda, my +1 as well.
    >     It is amazing to see the progress made in Submarine last year, the
    > community grows fast and quiet collaborative. I can see the reasons to get
    > it release faster in its own cycle. And at the same time, the Ozone way
    > works very well.
    >
    >     —
    >     Weiwei
    >     On Feb 1, 2019, 10:49 AM +0800, Xun Liu <[hidden email]>, wrote:
    >     > +1
    >     >
    >     > Hello everyone,
    >     >
    >     > I am Xun Liu, the head of the machine learning team at Netease
    > Research Institute. I quite agree with Wangda.
    >     >
    >     > Our team is very grateful for getting Submarine machine learning
    > engine from the community.
    >     > We are heavy users of Submarine.
    >     > Because Submarine fits into the direction of our big data team's
    > hadoop technology stack,
    >     > It avoids the needs to increase the manpower investment in learning
    > other container scheduling systems.
    >     > The important thing is that we can use a common YARN cluster to run
    > machine learning,
    >     > which makes the utilization of server resources more efficient, and
    > reserves a lot of human and material resources in our previous years.
    >     >
    >     > Our team have finished the test and deployment of the Submarine and
    > will provide the service to our e-commerce department (
    > http://www.kaola.com/) shortly.
    >     >
    >     > We also plan to provides the Submarine engine in our existing YARN
    > cluster in the next six months.
    >     > Because we have a lot of product departments need to use machine
    > learning services,
    >     > for example:
    >     > 1) Game department (http://game.163.com/) needs AI battle training,
    >     > 2) News department (http://www.163.com) needs news recommendation,
    >     > 3) Mailbox department (http://www.163.com) requires anti-spam and
    > illegal detection,
    >     > 4) Music department (https://music.163.com/) requires music
    > recommendation,
    >     > 5) Education department (http://www.youdao.com) requires voice
    > recognition,
    >     > 6) Massive Open Online Courses (https://open.163.com/) requires
    > multilingual translation and so on.
    >     >
    >     > If Submarine can be released independently like Ozone, it will help
    > us quickly get the latest features and improvements, and it will be great
    > helpful to our team and users.
    >     >
    >     > Thanks hadoop Community!
    >     >
    >     >
    >     > > 在 2019年2月1日,上午2:53,Wangda Tan <[hidden email]> 写道:
    >     > >
    >     > > Hi devs,
    >     > >
    >     > > Since we started submarine-related effort last year, we received a
    > lot of
    >     > > feedbacks, several companies (such as Netease, China Mobile, etc.)
    > are
    >     > > trying to deploy Submarine to their Hadoop cluster along with big
    > data
    >     > > workloads. Linkedin also has big interests to contribute a
    > Submarine TonY (
    >     > > https://github.com/linkedin/TonY) runtime to allow users to use
    > the same
    >     > > interface.
    >     > >
    >     > > From what I can see, there're several issues of putting Submarine
    > under
    >     > > yarn-applications directory and have same release cycle with
    > Hadoop:
    >     > >
    >     > > 1) We started 3.2.0 release at Sep 2018, but the release is done
    > at Jan
    >     > > 2019. Because of non-predictable blockers and security issues, it
    > got
    >     > > delayed a lot. We need to iterate submarine fast at this point.
    >     > >
    >     > > 2) We also see a lot of requirements to use Submarine on older
    > Hadoop
    >     > > releases such as 2.x. Many companies may not upgrade Hadoop to 3.x
    > in a
    >     > > short time, but the requirement to run deep learning is urgent to
    > them. We
    >     > > should decouple Submarine from Hadoop version.
    >     > >
    >     > > And why we wanna to keep it within Hadoop? First, Submarine
    > included some
    >     > > innovation parts such as enhancements of user experiences for YARN
    >     > > services/containerization support which we can add it back to
    > Hadoop later
    >     > > to address common requirements. In addition to that, we have a big
    > overlap
    >     > > in the community developing and using it.
    >     > >
    >     > > There're several proposals we have went through during Ozone merge
    > to trunk
    >     > > discussion:
    >     > >
    > https://mail-archives.apache.org/mod_mbox/hadoop-common-dev/201803.mbox/%3CCAHfHakH6_m3YLdf5a2KQ8+w-5fbVX5aHFgS-x1VaJW8gmnzRLg@...%3E
    >     > >
    >     > > I propose to adopt Ozone model: which is the same master branch,
    > different
    >     > > release cycle, and different release branch. It is a great example
    > to show
    >     > > agile release we can do (2 Ozone releases after Oct 2018) with less
    >     > > overhead to setup CI, projects, etc.
    >     > >
    >     > > *Links:*
    >     > > - JIRA: https://issues.apache.org/jira/browse/YARN-8135
    >     > > - Design doc
    >     > > <
    > https://docs.google.com/document/d/199J4pB3blqgV9SCNvBbTqkEoQdjoyGMjESV4MktCo0k/edit
    > >
    >     > > - User doc
    >     > > <
    > https://hadoop.apache.org/docs/r3.2.0/hadoop-yarn/hadoop-yarn-applications/hadoop-yarn-submarine/Index.html
    > >
    >     > > (3.2.0
    >     > > release)
    >     > > - Blogposts, {Submarine} : Running deep learning workloads on
    > Apache Hadoop
    >     > > <
    > https://hortonworks.com/blog/submarine-running-deep-learning-workloads-apache-hadoop/
    > >,
    >     > > (Chinese Translation: Link <https://www.jishuwen.com/d/2Vpu>)
    >     > > - Talks: Strata Data Conf NY
    >     > > <
    > https://conferences.oreilly.com/strata/strata-ny-2018/public/schedule/detail/68289
    > >
    >     > >
    >     > > Thoughts?
    >     > >
    >     > > Thanks,
    >     > > Wangda Tan
    >     >
    >     >
    >     >
    >     > ---------------------------------------------------------------------
    >     > To unsubscribe, e-mail: [hidden email]
    >     > For additional commands, e-mail: [hidden email]
    >     >
    >
    >
    >
   


---------------------------------------------------------------------
To unsubscribe, e-mail: [hidden email]
For additional commands, e-mail: [hidden email]
Reply | Threaded
Open this post in threaded view
|

Re: [DISCUSS] Making submarine to different release model like Ozone

Xiaoyu Yao
+1, thanks for bringing this up, Wangda. This will help expanding the Hadoop ecosystem by supporting new AI/ML workloads.

Thanks,
Xiaoyu
On 2/1/19, 10:58 AM, "Dinesh Chitlangia" <[hidden email]> wrote:

    +1 This is a fantastic recommendation given the increasing interest in ML across the globe.
   
    Thanks,
    Dinesh
   
   
   
    On 2/1/19, 1:54 PM, "Ajay Kumar" <[hidden email]> wrote:
   
        +1, Thanks for driving this. With rise of use cases running ML along with traditional applications this will be of great help.
       
        Thanks,
        Ajay  
       
        On 2/1/19, 10:49 AM, "Suma Shivaprasad" <[hidden email]> wrote:
       
            +1. Thanks for bringing this up Wangda.
           
            Makes sense to have Submarine follow its own release cadence given the good
            momentum/adoption so far. Also, making it run with older versions of Hadoop
            would drive higher adoption.
           
            Suma
           
            On Fri, Feb 1, 2019 at 9:40 AM Eric Yang <[hidden email]> wrote:
           
            > Submarine is an application built for YARN framework, but it does not have
            > strong dependency on YARN development.  For this kind of projects, it would
            > be best to enter Apache Incubator cycles to create a new community.  Apache
            > commons is the only project other than Incubator that has independent
            > release cycles.  The collection is large, and the project goal is
            > ambitious.  No one really knows which component works with each other in
            > Apache commons.  Hadoop is a much more focused project on distributed
            > computing framework and not incubation sandbox.  For alignment with Hadoop
            > goals, and we want to prevent Hadoop project to be overloaded while
            > allowing good ideas to be carried forwarded in Apache incubator.  Put on my
            > Apache Member hat, my vote is -1 to allow more independent subproject
            > release cycle in Hadoop project that does not align with Hadoop project
            > goals.
            >
            > Apache incubator process is highly recommended for Submarine:
            > https://incubator.apache.org/policy/process.html This allows Submarine to
            > develop for older version of Hadoop like Spark works with multiple versions
            > of Hadoop.
            >
            > Regards,
            > Eric
            >
            > On 1/31/19, 10:51 PM, "Weiwei Yang" <[hidden email]> wrote:
            >
            >     Thanks for proposing this Wangda, my +1 as well.
            >     It is amazing to see the progress made in Submarine last year, the
            > community grows fast and quiet collaborative. I can see the reasons to get
            > it release faster in its own cycle. And at the same time, the Ozone way
            > works very well.
            >
            >     —
            >     Weiwei
            >     On Feb 1, 2019, 10:49 AM +0800, Xun Liu <[hidden email]>, wrote:
            >     > +1
            >     >
            >     > Hello everyone,
            >     >
            >     > I am Xun Liu, the head of the machine learning team at Netease
            > Research Institute. I quite agree with Wangda.
            >     >
            >     > Our team is very grateful for getting Submarine machine learning
            > engine from the community.
            >     > We are heavy users of Submarine.
            >     > Because Submarine fits into the direction of our big data team's
            > hadoop technology stack,
            >     > It avoids the needs to increase the manpower investment in learning
            > other container scheduling systems.
            >     > The important thing is that we can use a common YARN cluster to run
            > machine learning,
            >     > which makes the utilization of server resources more efficient, and
            > reserves a lot of human and material resources in our previous years.
            >     >
            >     > Our team have finished the test and deployment of the Submarine and
            > will provide the service to our e-commerce department (
            > http://www.kaola.com/) shortly.
            >     >
            >     > We also plan to provides the Submarine engine in our existing YARN
            > cluster in the next six months.
            >     > Because we have a lot of product departments need to use machine
            > learning services,
            >     > for example:
            >     > 1) Game department (http://game.163.com/) needs AI battle training,
            >     > 2) News department (http://www.163.com) needs news recommendation,
            >     > 3) Mailbox department (http://www.163.com) requires anti-spam and
            > illegal detection,
            >     > 4) Music department (https://music.163.com/) requires music
            > recommendation,
            >     > 5) Education department (http://www.youdao.com) requires voice
            > recognition,
            >     > 6) Massive Open Online Courses (https://open.163.com/) requires
            > multilingual translation and so on.
            >     >
            >     > If Submarine can be released independently like Ozone, it will help
            > us quickly get the latest features and improvements, and it will be great
            > helpful to our team and users.
            >     >
            >     > Thanks hadoop Community!
            >     >
            >     >
            >     > > 在 2019年2月1日,上午2:53,Wangda Tan <[hidden email]> 写道:
            >     > >
            >     > > Hi devs,
            >     > >
            >     > > Since we started submarine-related effort last year, we received a
            > lot of
            >     > > feedbacks, several companies (such as Netease, China Mobile, etc.)
            > are
            >     > > trying to deploy Submarine to their Hadoop cluster along with big
            > data
            >     > > workloads. Linkedin also has big interests to contribute a
            > Submarine TonY (
            >     > > https://github.com/linkedin/TonY) runtime to allow users to use
            > the same
            >     > > interface.
            >     > >
            >     > > From what I can see, there're several issues of putting Submarine
            > under
            >     > > yarn-applications directory and have same release cycle with
            > Hadoop:
            >     > >
            >     > > 1) We started 3.2.0 release at Sep 2018, but the release is done
            > at Jan
            >     > > 2019. Because of non-predictable blockers and security issues, it
            > got
            >     > > delayed a lot. We need to iterate submarine fast at this point.
            >     > >
            >     > > 2) We also see a lot of requirements to use Submarine on older
            > Hadoop
            >     > > releases such as 2.x. Many companies may not upgrade Hadoop to 3.x
            > in a
            >     > > short time, but the requirement to run deep learning is urgent to
            > them. We
            >     > > should decouple Submarine from Hadoop version.
            >     > >
            >     > > And why we wanna to keep it within Hadoop? First, Submarine
            > included some
            >     > > innovation parts such as enhancements of user experiences for YARN
            >     > > services/containerization support which we can add it back to
            > Hadoop later
            >     > > to address common requirements. In addition to that, we have a big
            > overlap
            >     > > in the community developing and using it.
            >     > >
            >     > > There're several proposals we have went through during Ozone merge
            > to trunk
            >     > > discussion:
            >     > >
            > https://mail-archives.apache.org/mod_mbox/hadoop-common-dev/201803.mbox/%3CCAHfHakH6_m3YLdf5a2KQ8+w-5fbVX5aHFgS-x1VaJW8gmnzRLg@...%3E
            >     > >
            >     > > I propose to adopt Ozone model: which is the same master branch,
            > different
            >     > > release cycle, and different release branch. It is a great example
            > to show
            >     > > agile release we can do (2 Ozone releases after Oct 2018) with less
            >     > > overhead to setup CI, projects, etc.
            >     > >
            >     > > *Links:*
            >     > > - JIRA: https://issues.apache.org/jira/browse/YARN-8135
            >     > > - Design doc
            >     > > <
            > https://docs.google.com/document/d/199J4pB3blqgV9SCNvBbTqkEoQdjoyGMjESV4MktCo0k/edit
            > >
            >     > > - User doc
            >     > > <
            > https://hadoop.apache.org/docs/r3.2.0/hadoop-yarn/hadoop-yarn-applications/hadoop-yarn-submarine/Index.html
            > >
            >     > > (3.2.0
            >     > > release)
            >     > > - Blogposts, {Submarine} : Running deep learning workloads on
            > Apache Hadoop
            >     > > <
            > https://hortonworks.com/blog/submarine-running-deep-learning-workloads-apache-hadoop/
            > >,
            >     > > (Chinese Translation: Link <https://www.jishuwen.com/d/2Vpu>)
            >     > > - Talks: Strata Data Conf NY
            >     > > <
            > https://conferences.oreilly.com/strata/strata-ny-2018/public/schedule/detail/68289
            > >
            >     > >
            >     > > Thoughts?
            >     > >
            >     > > Thanks,
            >     > > Wangda Tan
            >     >
            >     >
            >     >
            >     > ---------------------------------------------------------------------
            >     > To unsubscribe, e-mail: [hidden email]
            >     > For additional commands, e-mail: [hidden email]
            >     >
            >
            >
            >
           
       
       
        ---------------------------------------------------------------------
        To unsubscribe, e-mail: [hidden email]
        For additional commands, e-mail: [hidden email]
       
       
   
   
    ---------------------------------------------------------------------
    To unsubscribe, e-mail: [hidden email]
    For additional commands, e-mail: [hidden email]
   
   


---------------------------------------------------------------------
To unsubscribe, e-mail: [hidden email]
For additional commands, e-mail: [hidden email]
Reply | Threaded
Open this post in threaded view
|

Re: [DISCUSS] Making submarine to different release model like Ozone

Wangda Tan
In reply to this post by John Zhuge-2
Eric,
Thanks for your reconsideration. We will definitely try best to not break
compatibilities, etc. like how we did to other components!

Really appreciate everybody's support, thoughts, suggestions shared on this
thread. Given the discussion went very positive, I will go ahead to send a
voting thread.

Best,
Wangda

On Fri, Feb 1, 2019 at 2:06 PM Eric Yang <[hidden email]> wrote:

> If HDFS or YARN breaks compatibility with Submarine, it will require to
> make release to catch up with the latest Hadoop changes.  On
> hadoop.apache.org website, the latest news may always have Submarine on
> top to repair compatibility with latest of Hadoop.  This may overwhelm any
> interesting news that may happen in Hadoop space.  I don’t like to see that
> happen, but unavoidable with independent release cycle.  Maybe there is a
> good way to avoid this with help of release manager to ensure that
> Hadoop/Submarine don’t break compatibility frequently.
>
>
>
> For me to lift my veto, release managers of independent release cycles
> need to take responsibility to ensure X version of Hadoop is tested with Y
> version of Submarine.  Release managers will have to do more work to ensure
> the defined combination works.  With the greater responsibility of release
> management comes with its own reward.  Seasoned PMC may be nominated to
> become Apache Member, which will help with Submarine to enter Apache
> Incubator when time is right.  Hence, I will withdraw my veto and let
> Submarine set its own course.
>
>
>
> Good luck Wangda.
>
>
>
> Regards,
>
> Eric
>
>
>
> *From: *Wangda Tan <[hidden email]>
> *Date: *Friday, February 1, 2019 at 10:52 AM
> *To: *Eric Yang <[hidden email]>
> *Cc: *Weiwei Yang <[hidden email]>, Xun Liu <[hidden email]>,
> Hadoop Common <[hidden email]>, "[hidden email]"
> <[hidden email]>, Hdfs-dev <[hidden email]>, "
> [hidden email]" <[hidden email]>
> *Subject: *Re: [DISCUSS] Making submarine to different release model like
> Ozone
>
>
>
> Thanks everyone for sharing thoughts!
>
>
>
> Eric, appreciate your suggestions. But there are many examples to have
> separate releases, like Hive's storage API, OZone, etc. For loosely coupled
> sub-projects, it gonna be great (at least for most of the users) to have
> separate releases so new features can be faster consumed and iterated. From
> above feedbacks from developers and users, I think it is also what people
> want.
>
>
>
> Another concern you mentioned is Submarine is aligned with Hadoop project
> goals. From feedbacks we can see, it attracts companies continue using
> Hadoop to solve their ML/DL requirements, it also created a good feedback
> loop, many issues faced, and some new functionalities added by Submarine
> went back to Hadoop. Such as localization files, directories. GPU topology
> related enhancement, etc.
>
>
>
> We will definitely use this sub-project opportunity to fast grow both
> Submarine and Hadoop, try to get fast release cycles for both of the
> projects. And for your suggestion about Apache incubator, we can reconsider
> it once Submarine becomes a more independent project, now it is still too
> small and too much overhead to go through the process, I don't want to stop
> the fast-growing community for months to go through incubator process for
> now.
>
>
>
> I really hope my comment can help you reconsider the veto. :)
>
>
>
> Thanks,
>
> Wangda
>
>
>
> On Fri, Feb 1, 2019 at 9:39 AM Eric Yang <[hidden email]> wrote:
>
> Submarine is an application built for YARN framework, but it does not have
> strong dependency on YARN development.  For this kind of projects, it would
> be best to enter Apache Incubator cycles to create a new community.  Apache
> commons is the only project other than Incubator that has independent
> release cycles.  The collection is large, and the project goal is
> ambitious.  No one really knows which component works with each other in
> Apache commons.  Hadoop is a much more focused project on distributed
> computing framework and not incubation sandbox.  For alignment with Hadoop
> goals, and we want to prevent Hadoop project to be overloaded while
> allowing good ideas to be carried forwarded in Apache incubator.  Put on my
> Apache Member hat, my vote is -1 to allow more independent subproject
> release cycle in Hadoop project that does not align with Hadoop project
> goals.
>
> Apache incubator process is highly recommended for Submarine:
> https://incubator.apache.org/policy/process.html This allows Submarine to
> develop for older version of Hadoop like Spark works with multiple versions
> of Hadoop.
>
> Regards,
> Eric
>
> On 1/31/19, 10:51 PM, "Weiwei Yang" <[hidden email]> wrote:
>
>     Thanks for proposing this Wangda, my +1 as well.
>     It is amazing to see the progress made in Submarine last year, the
> community grows fast and quiet collaborative. I can see the reasons to get
> it release faster in its own cycle. And at the same time, the Ozone way
> works very well.
>
>     —
>     Weiwei
>     On Feb 1, 2019, 10:49 AM +0800, Xun Liu <[hidden email]>, wrote:
>     > +1
>     >
>     > Hello everyone,
>     >
>     > I am Xun Liu, the head of the machine learning team at Netease
> Research Institute. I quite agree with Wangda.
>     >
>     > Our team is very grateful for getting Submarine machine learning
> engine from the community.
>     > We are heavy users of Submarine.
>     > Because Submarine fits into the direction of our big data team's
> hadoop technology stack,
>     > It avoids the needs to increase the manpower investment in learning
> other container scheduling systems.
>     > The important thing is that we can use a common YARN cluster to run
> machine learning,
>     > which makes the utilization of server resources more efficient, and
> reserves a lot of human and material resources in our previous years.
>     >
>     > Our team have finished the test and deployment of the Submarine and
> will provide the service to our e-commerce department (
> http://www.kaola.com/) shortly.
>     >
>     > We also plan to provides the Submarine engine in our existing YARN
> cluster in the next six months.
>     > Because we have a lot of product departments need to use machine
> learning services,
>     > for example:
>     > 1) Game department (http://game.163.com/) needs AI battle training,
>     > 2) News department (http://www.163.com) needs news recommendation,
>     > 3) Mailbox department (http://www.163.com) requires anti-spam and
> illegal detection,
>     > 4) Music department (https://music.163.com/) requires music
> recommendation,
>     > 5) Education department (http://www.youdao.com) requires voice
> recognition,
>     > 6) Massive Open Online Courses (https://open.163.com/) requires
> multilingual translation and so on.
>     >
>     > If Submarine can be released independently like Ozone, it will help
> us quickly get the latest features and improvements, and it will be great
> helpful to our team and users.
>     >
>     > Thanks hadoop Community!
>     >
>     >
>     > > 在 2019年2月1日,上午2:53,Wangda Tan <[hidden email]> 写道:
>     > >
>     > > Hi devs,
>     > >
>     > > Since we started submarine-related effort last year, we received a
> lot of
>     > > feedbacks, several companies (such as Netease, China Mobile, etc.)
> are
>     > > trying to deploy Submarine to their Hadoop cluster along with big
> data
>     > > workloads. Linkedin also has big interests to contribute a
> Submarine TonY (
>     > > https://github.com/linkedin/TonY) runtime to allow users to use
> the same
>     > > interface.
>     > >
>     > > From what I can see, there're several issues of putting Submarine
> under
>     > > yarn-applications directory and have same release cycle with
> Hadoop:
>     > >
>     > > 1) We started 3.2.0 release at Sep 2018, but the release is done
> at Jan
>     > > 2019. Because of non-predictable blockers and security issues, it
> got
>     > > delayed a lot. We need to iterate submarine fast at this point.
>     > >
>     > > 2) We also see a lot of requirements to use Submarine on older
> Hadoop
>     > > releases such as 2.x. Many companies may not upgrade Hadoop to 3.x
> in a
>     > > short time, but the requirement to run deep learning is urgent to
> them. We
>     > > should decouple Submarine from Hadoop version.
>     > >
>     > > And why we wanna to keep it within Hadoop? First, Submarine
> included some
>     > > innovation parts such as enhancements of user experiences for YARN
>     > > services/containerization support which we can add it back to
> Hadoop later
>     > > to address common requirements. In addition to that, we have a big
> overlap
>     > > in the community developing and using it.
>     > >
>     > > There're several proposals we have went through during Ozone merge
> to trunk
>     > > discussion:
>     > >
> https://mail-archives.apache.org/mod_mbox/hadoop-common-dev/201803.mbox/%3CCAHfHakH6_m3YLdf5a2KQ8+w-5fbVX5aHFgS-x1VaJW8gmnzRLg@...%3E
>     > >
>     > > I propose to adopt Ozone model: which is the same master branch,
> different
>     > > release cycle, and different release branch. It is a great example
> to show
>     > > agile release we can do (2 Ozone releases after Oct 2018) with less
>     > > overhead to setup CI, projects, etc.
>     > >
>     > > *Links:*
>     > > - JIRA: https://issues.apache.org/jira/browse/YARN-8135
>     > > - Design doc
>     > > <
> https://docs.google.com/document/d/199J4pB3blqgV9SCNvBbTqkEoQdjoyGMjESV4MktCo0k/edit
> >
>     > > - User doc
>     > > <
> https://hadoop.apache.org/docs/r3.2.0/hadoop-yarn/hadoop-yarn-applications/hadoop-yarn-submarine/Index.html
> >
>     > > (3.2.0
>     > > release)
>     > > - Blogposts, {Submarine} : Running deep learning workloads on
> Apache Hadoop
>     > > <
> https://hortonworks.com/blog/submarine-running-deep-learning-workloads-apache-hadoop/
> >,
>     > > (Chinese Translation: Link <https://www.jishuwen.com/d/2Vpu>)
>     > > - Talks: Strata Data Conf NY
>     > > <
> https://conferences.oreilly.com/strata/strata-ny-2018/public/schedule/detail/68289
> >
>     > >
>     > > Thoughts?
>     > >
>     > > Thanks,
>     > > Wangda Tan
>     >
>     >
>     >
>     > ---------------------------------------------------------------------
>     > To unsubscribe, e-mail: [hidden email]
>     > For additional commands, e-mail: [hidden email]
>     >
>
>