Hi!
I feel very excited about this idea. From what I can see, the community has
acted very mature. Releases and community growth are happening as well. The
community is diversifying and the project's life and death no longer hinges
on single individuals.
Stephan
Am 28.11.2014 01:27 schrieb "Fabia
I agree with Kostas.
Let's go for the graduation!
Cheers, Fabian
2014-11-27 20:34 GMT+01:00 Kostas Tzoumas :
> Hi everyone,
>
> I would like to share a discussion that we have been having on the PPMC
> list.
>
> Flink has successfully made two Apache releases and added 5 committers
> since enter
Thanks for putting this together. Looks really good!
IMO, the tag line should make clear that Flink is a system for parallel
data processing.
2014-11-27 23:35 GMT+01:00 Markl, Volker, Prof. Dr. <
volker.ma...@tu-berlin.de>:
> +1
>
> Von meinem iPhone gesendet
>
> > Am 27.11.2014 um 20:37 schrie
+1
Von meinem iPhone gesendet
> Am 27.11.2014 um 20:37 schrieb "Gyula Fora" :
>
> +1
>
> I also think that at looks very good!
> Stylish :)
>
>> On 27 Nov 2014, at 20:32, Stephan Ewen wrote:
>>
>> I like it as well. Much better than the current website :-)
>>
>> On Thu, Nov 27, 2014 at 8:30
I also like the shorter tag line.
What about using 3 attributes.. As three is a good number and it also
preserves the layout that the text is beneath and both sides of the logo
Like
Fast, reliable, large scale data processing.
Or with clusters.. It should be at least "on clusters" 😁
Johanne
I also like the design.. Looks almost too good for an open source project 😁
Some minor things
Shouldn't it be run on Yarn as contrary to run in Yarn?
Also the section about reliable and scalable. This could benefit from a bit of
rewording.. Guess all three paragraphs should mention the word FL
+1
The new website presents a much better overview for the general user.
On Thu, Nov 27, 2014 at 8:41 PM, Vasiliki Kalavri wrote:
> Hey!
>
> It looks great :))
>
> If I would only change one thing, I would remove "in clusters" from the
> tagline.
> First, because it can also be locally and in t
Hey!
It looks great :))
If I would only change one thing, I would remove "in clusters" from the
tagline.
First, because it can also be locally and in the cloud and second, because
the shorter the tagline the better, no? ;)
Cheers,
V.
On 27 November 2014 at 20:34, Gyula Fora wrote:
> +1
>
> I
+1
I also think that at looks very good!
Stylish :)
> On 27 Nov 2014, at 20:32, Stephan Ewen wrote:
>
> I like it as well. Much better than the current website :-)
>
> On Thu, Nov 27, 2014 at 8:30 PM, Aljoscha Krettek
> wrote:
>
>> The formatting of the Scala example code is a bit off in pla
Hi everyone,
I would like to share a discussion that we have been having on the PPMC
list.
Flink has successfully made two Apache releases and added 5 committers
since entering the Incubator. The community is openly discussing in the
mailing lists, and new contributors and users are coming to the
I like it as well. Much better than the current website :-)
On Thu, Nov 27, 2014 at 8:30 PM, Aljoscha Krettek
wrote:
> The formatting of the Scala example code is a bit off in places. Other
> than that it looks very good.
>
> On Thu, Nov 27, 2014 at 8:22 PM, Kostas Tzoumas
> wrote:
> > Actually
The formatting of the Scala example code is a bit off in places. Other
than that it looks very good.
On Thu, Nov 27, 2014 at 8:22 PM, Kostas Tzoumas wrote:
> Actually, I cancelled the commit.
>
> Perhaps this is a good time to comment on the new frontpage (the rest of
> the content is the same).
Stephan Ewen created FLINK-1292:
---
Summary: Allow for longer normalized keys
Key: FLINK-1292
URL: https://issues.apache.org/jira/browse/FLINK-1292
Project: Flink
Issue Type: Improvement
Stephan Ewen created FLINK-1291:
---
Summary: Remove GC options from JobManager and TaskManager scripts
Key: FLINK-1291
URL: https://issues.apache.org/jira/browse/FLINK-1291
Project: Flink
Issue T
Actually, I cancelled the commit.
Perhaps this is a good time to comment on the new frontpage (the rest of
the content is the same).
You can see the frontage in pdf (two parts, as you would scroll down) here:
https://www.dropbox.com/sh/4b9f4judlczfzkp/AADIsdw1vxyNVZH6oyWsAIWNa?dl=0
I think we ca
Hi everyone,
Ufuk, Robert, and I worked on a new layout for the Flink website, and added
some content to the frontpage to (hopefully) convey more clearly what Flink
is all about and what are Flink's differentiating features (a point of
criticism to the old website was often that it is very non-des
Hey,
Thanks Vasia, great idea.
When we've started the streaming development the first sketch API we had is
was actually really similar to Storm's API a couple of versions back. I've
talked to Gyula about the issue when we've first heard about SAMOA ago and
we are confident that porting Samoa to Fl
Hey!
As an example, you can have a look at the normalized key sorter, which
sorts and compares complete on serialized data.
Stephan
On Thu, Nov 27, 2014 at 2:25 PM, Viktor Rosenfeld <
viktor.rosenf...@tu-berlin.de> wrote:
> Hi Stephan,
>
> can you give me an example (or a few) where Flink is w
Hi Stephan,
can you give me an example (or a few) where Flink is working on the
serialized data?
Cheers,
Viktor
--
View this message in context:
http://apache-flink-incubator-mailing-list-archive.1008284.n3.nabble.com/Compression-of-network-traffic-tp2568p2651.html
Sent from the Apache Flink
Hey Vasia,
I really like the idea and it seems we have some volunteers from our side
already!
It is in my opinion a starter task since it requires to create a simple adapter
for flink as well as mapping abstractions to our streaming topology API in a
similar fashion (in its simplest form just u
Hello everyone,
Last week at Strata, I attended a talk about a streaming machine learning
framework from Yahoo labs, SAMOA: https://github.com/yahoo/samoa. The
project is going through apache incubation application and currently runs
on Storm, S4 and Samza.
After the talk, I spoke with the guy th
Stephan Ewen created FLINK-1290:
---
Summary: Optimizer prunes all candidates when unable to reuse sort
properties
Key: FLINK-1290
URL: https://issues.apache.org/jira/browse/FLINK-1290
Project: Flink
Ahh, I didn't see that. My bad.
On Thu, Nov 27, 2014 at 11:47 AM, Fabian Hueske wrote:
> Viktor said he changed the implementation to
> MapFunction -> ReduceFunction -> MapFunction.
>
> So it is combinable :-)
>
> 2014-11-27 11:45 GMT+01:00 Aljoscha Krettek :
>
>> Hi,
>> why does the GroupReduce
Hi Viktor,
thanks for the update!
Regarding the explicit vs implicit adding of key fields:
I would only allow to use either key(x) or allKeys() and throw an exception
if they are mixed.
I guess there won't be many situations, where somebody would want to mix
them anyway.
No need to complicate the
Viktor said he changed the implementation to
MapFunction -> ReduceFunction -> MapFunction.
So it is combinable :-)
2014-11-27 11:45 GMT+01:00 Aljoscha Krettek :
> Hi,
> why does the GroupReduce change the output type? Can this not be done
> in the two mappers? In my opinion, aggregations should
Hi,
why does the GroupReduce change the output type? Can this not be done
in the two mappers? In my opinion, aggregations should be combinable,
otherwise, performance would be severely crippled.
Cheers,
Aljoscha
On Thu, Nov 27, 2014 at 11:20 AM, Viktor Rosenfeld
wrote:
> Hi Fabian,
>
> thanks fo
Hi Fabian,
thanks for your feedback. See my responses below.
Fabian Hueske wrote
> - I would split the branch into two branches, one for each approach. That
> make comparisons with master much easier.
I've moved the changes necessary for the second approach to a branch called
aggregation-alt:
h
Our implication so far was that forwarding means evenly scattering over
successors - a balanced load being the important goal.
If you find different requirements in streaming, you could define a new
type of selector.
On Thu, Nov 27, 2014 at 11:02 AM, Gyula Fora wrote:
> Thanks Stephan,
>
> So I
Thanks Stephan,
So I took a quick look at the ChannelSelectors the batch api uses and I see
that for Forward strategy uses round-robin. My question was aimed exactly to
avoid having to do this. Isn’t this sub-optimal?
Maybe we could pass the channel info to the channel selector, so it can make
This is a bit tricky, since the new scheduling is more flexible...
Assume we have a PointWise connection with two receiving tasks per sending
task: outgoing channels 0 and 1.
When scheduling in the most basic mode, the receivers can go anywhere, but
the schedule will try to give them a slot on th
Yes, that working on serialized data happens in parts right now and it
would be great to extend that.
While it would be possible to work on a compact serialized representation,
I can't think of a way to work on a snappy/lz4 compressed version.
Am 26.11.2014 23:00 schrieb "Viktor Rosenfeld" <
vikto
31 matches
Mail list logo
