[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Beam Samza Runner status update

I'm excited to hear about making Samza capable of running non-Java pipelines.

On Fri, Oct 12, 2018 at 9:46 AM Xinyu Liu <xinyuliu.us@xxxxxxxxx> wrote:
@Max: absolutely we should work together! FlinkRunner has been our best reference since the start of our SamzaRunner, and the previous work in Flink portable runner has been extremely valuable to us too. We haven't got to the point of portable stateful processing yet. Our next step is to hook up a streaming source, i.e. Kafka, and test out streaming capabilities such as watermarks, windowing and triggers. For us, reading from Kafka will happen on the Java side (we have quite a lot of extensions of Kafka in LinkedIn), so we will try to create some internal Python API and do the translation right now. On the other hand, we are following up with BEAM-2937 for the optimization of Combine in portable runner. Previously we run into some state problem without this being resolved. Anyway, look forward to syncing up with you more!


On Fri, Oct 12, 2018 at 1:40 AM Maximilian Michels <mxm@xxxxxxxxxx> wrote:
Thanks for the updating, Xinyu and Hai! Great to see another Running
emerging :)

I'm on the FlinkRunner. Looking forward to working together with you to
make the Beam Runners even better. Particularly, we should sync on the
portability, as some things are still to be fleshed out. In Flink, we
are starting to integrate portable State.


On 11.10.18 05:14, Jesse Anderson wrote:
> Interesting
> On Wed, Oct 10, 2018, 3:49 PM Kenneth Knowles <kenn@xxxxxxxxxx
> <mailto:kenn@xxxxxxxxxx>> wrote:
>     Welcome, Hai!
>     On Wed, Oct 10, 2018 at 3:46 PM Hai Lu <lhaiesp@xxxxxxxxx
>     <mailto:lhaiesp@xxxxxxxxx>> wrote:
>         Hi, all
>         This is Hai from LinkedIn. As Xinyu mentioned, I have been
>         working on portable API for Samza runner and made some solid
>         progress. It's been a very smooth process (although not
>         effortless for sure) and I'm really grateful for the great
>         platform that you all have built. I'm very impressed. Bravo!
>         Excited to work with everyone on Beam. Do expect more questions
>         from me down the road.
>         Thanks,
>         Hai
>         On Wed, Oct 10, 2018 at 12:36 PM Kenneth Knowles
>         <kenn@xxxxxxxxxx <mailto:kenn@xxxxxxxxxx>> wrote:
>             Clarification: Thomas Groh wrote the fuser, not me!
>             Thanks for the sharing all this. Really cool.
>             Kenn
>             On Wed, Oct 10, 2018 at 11:17 AM Rui Wang <ruwang@xxxxxxxxxx
>             <mailto:ruwang@xxxxxxxxxx>> wrote:
>                 Thanks for sharing! it's so exciting to hear that Beam
>                 is being used on Samza in production @LinkedIn! Your
>                 feedback will be helpful to Beam community!
>                 Besides, Beam supports SQL right now and hopefully Beam
>                 community could also receive feedback on BeamSQL
>                 <https://beam.apache.org/documentation/dsls/sql/overview/> in
>                 the future.
>                 -Rui
>                 On Wed, Oct 10, 2018 at 11:10 AM Jean-Baptiste Onofré
>                 <jb@xxxxxxxxxxxx <mailto:jb@xxxxxxxxxxxx>> wrote:
>                     Thanks for sharing and congrats for this great work !
>                     Regards
>                     JB
>                     Le 10 oct. 2018, à 20:23, Xinyu Liu <xinyuliu.us
>                     <mailto:%3Ca>@gmail.com <http://gmail.com>
>                     target=_blank>xinyuliu.us
>                     <http://xinyuliu.us>@gmail.com <http://gmail.com>> a
>                     écrit:
>                         Hi, All,
>                         It's been over four months since we added the
>                         Samza Runner to Beam, and we've been making a
>                         lot of progress after that. Here I would like to
>                         update your guys and share some really good news
>                         happening here at LinkedIn:
>                         1) First Beam job in production @LInkedIn!
>                         After a few rounds of testing and benchmarking,
>                         we finally rolled out our first Beam job here!
>                         The job uses quite a few features, such as event
>                         time, fixed/session windowing, early triggering,
>                         and stateful processing. Our first customer is
>                         very happy and they highly appraise the
>                         easy-to-use Beam API as well as powerful
>                         processing model. Due to the limited resources
>                         here, we put our full trust in the work you guys
>                         are doing, and we didn't run into any surprises.
>                         We see extremely attention to details as well as
>                         non-compromise in any user experience everywhere
>                         in the code base. We would like to thank
>                         everyone in the Beam community to contribute to
>                         such an amazing framework!
>                         2) A portable Samza Runner prototype
>                         We are also starting the work in making Samza
>                         Runner portable. So far we just got the python
>                         word count example working using portable Samza
>                         Runner. Please look out for the PR for this very
>                         soon :). Again, this work is not possible
>                         without the great Beam portability framework,
>                         and the developers like Luke and Ahmet, just to
>                         name a few, behind it. The ReferenceRunner has
>                         been extremely useful to us to figure out what's
>                         needed and how it works. Kudos to Thomas Groh,
>                         Ben Sidhom and all the others who makes this
>                         available to us. And to Kenn, your fuse work rocks.
>                         3) More contributors in Samza Runner
>                         The runner has been Chris and my personal
>                         project for a while and now it's not the case.
>                         We got Hai Lu and Boris Shkolnik from Samza team
>                         to contribute. Hai has been focusing on the
>                         portability work as mentioned in #2, and Boris
>                         will work mostly on supporting our use cases. We
>                         will send more emails discussing our use cases,
>                         like the "Update state after firing" email I
>                         sent out earlier.
>                         Finally, a shout-out to our very own Chris
>                         Pettitt. Without you, none of the above won't
>                         happen!
>                         Thanks,
>                         Xinyu