incubator-general mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From vongosling <fengji...@gmail.com>
Subject Re: [DISCUSS] RocketMQ Incubation Proposal
Date Sat, 05 Nov 2016 10:14:57 GMT
Hi,John,Roman:

I am vongosling. The creator of RocketMQ. Please let me, on behalf of the
RocketMQ project, to talk about a period of history. May be it can explain
why we wrote "RocketMQ has open source since 2012". Actually, RocketMQ is
the third generation Distributed MOM product within alibaba. The other 2
generation products, we call it "MetaQ". So you can think it  we created
the MetaQ 1.0 and MetaQ 2.0.
The biggest luminescent spot of MetaQ1 is its scalable text protocol and
JTA support. In MetaQ 2, we let one broker support more partitions and less
IO resources. MetaQ1 has open source since 2012. About this product , we
can see from this link here https://github.com/killme2008/Metamorphosis.
Maybe the follwing chinese texts will give us some clue.

"MetaQ的淘宝版本已经发展到2.0乃至3.0版本,设计原则发生了变更,有兴趣的可以上它的官方网站看。”

The text above definitely showing that MetaQ has developed to 2.0 and 3.0.
No problem. In 3.0 version, we can not follow the previous name convention.
We renamed it RocketMQ. hoping it will be as fast as Rocket, low latency in
our high frequency trading scenario.Haha, yeah.Even now, MetaQ is still
maintained by some alibaba‘s tech. enthusiast in Github.

Like Roman said, "commit early, commit often" mentality, is always our open
source belief. According to what I said before, we really hope to find a
way to keep the same deploy, especially commit pace between Git
repositories. Thanks Roman’s advice 😊


 "there wasn't a ton of open source development on the product ". Thank's
john question. Please believe our open source mentality. In past 7 years,
With the alibaba business dramatical growth, our team faced a lot of
system architecture
and design challenges, we want to share our experience with every people.
But, We also hope more people to join us, to help us to perfect messaging
engine, which can bearing Trillion message  communication in the double
eleven Shopping Festival every year. For this purpose, we open a survey for
our customer as the link(https://github.com/alibaba/RocketMQ/issues/1 and
https://github.com/alibaba/RocketMQ/issues/353)  shows.Knowing your
customers, assembleing them together and making our community more health,
more active~

  "relationship with other apache products".There is no doubting, apache is
the most famous open source organization. Many companies, especially those
tech. companies, having benefited from apache's a ton of excellent open
source projects. Alibaba is also One of the beneficiaries, RocketMQ used a
lot of Apache Commons tools  in its every component.Also, this reason drive
us to donate our product. As an old Chinese saying goes  “What is taken
from the people is used in the interests of the people”. Here, we can
replace people with apache hah~

Also, IMHO, ActiveMQ, Kafka, RocektMQ and other apache MQ products. The
relationship, like apache’s many
streaming solutions(spark,flink,samza,storm etc.), they are
not competitors, They both have their own designs and do well in their
dedicated directions. BTW, within alibaba, in some bigdata departments,
Kafka is still their first choice. Also this reason, in my team, we have
some kafka tech. staffs to service them passionately.

 Last. Nowadays, RocketMQ storage is custom-built for low-latency purpose,
using JDK primitive FIle API and some JNI tech.. Different with activeMQ.
we are not support many storage strategies now. but we are planning to
extract storage layer. May be a  thin abstract layer, many concrete
implementation adapter (leverage Cassandra, Derby  or some off heap storage
engine). would occur. As another question, "cross platform
clients cooperativity". as I said above, some staffs in my team, are
dedicated to support service for kafka. they are developing anther product
kafka-proxy, make use of kafka existence client. The full data streaming
process is like this "kafka producer -> kafka proxy-> kafka consumer or
rocketmq consumer " Moreover, we are absorbing community proactive guys, to
assist us to complete cross language sdk like our design showing(
https://github.com/alibaba/RocketMQ/wiki/multi-language-sdk).

As proposal said, the ASF is the natural choice to host the RocketMQ
project as its goal of encouraging community-driven open source projects
fits with our vision for RocketMQ.

That's all, Thanks for everyone‘s advice and questions.



Best regards ~

from vongosling

2016-11-05 5:26 GMT+08:00 John D. Ament <john.d.ament@gmail.com>:

> On Fri, Nov 4, 2016 at 4:43 PM Roman Shaposhnik <roman@shaposhnik.org>
> wrote:
>
> > The proposal looks fine in general, but I'm slightly concerned about:
> >    https://github.com/alibaba/RocketMQ/graphs/contributors
> >
> > It seems that the model so far has been -- through huge blobs of
> > code over the wall. Given that the composition of initial committers
> > is all from Alibaba I hope their mentors will spend a lot of time
> > making sure that "commit early, commit often" mentality prevails.
> >
> > In addition to that, I can't seem to reconcile the statement:
> >    "The source code was opened up in 2012."
> > with what I see on GitHub. What am I missing?
> >
>
> So I think these are the same points I was bringing up as well.  I suspect
> its a case where there wasn't a ton of open source development on the
> product and it was kept internal.
>
> I'm still a bit leary about the "relationship with other apache products"
> section still.  I'm not interested in seeing how a podling competes with
> other projects (and its a bit surprising, since Bruce is the chair of one
> of the competitors), but instead how the podling has synergies with the
> other components.  I raised that they're using ASF projects today in their
> code base.
>
> Some other ways to address this section:
>
> - How can RocketMQ work with the existing Kafka or ActiveMQ communities to
> build cross platform clients?
> - How can RocketMQ look to leverage Cassandra, Geode, Derby as backend
> persistence stores?
>
> etc..
>
>
> >
> > Thanks,
> > Roman.
> >
> > On Fri, Nov 4, 2016 at 11:00 AM, Brian McCallister <brianm@skife.org>
> > wrote:
> > > +1 !
> > >
> > > On Fri, Nov 4, 2016 at 8:37 AM, Jim Jagielski <jim@jagunet.com> wrote:
> > >
> > >> Cool.
> > >>
> > >> +1
> > >>
> > >> > On Nov 3, 2016, at 6:10 PM, Bruce Snyder <bruce.snyder@gmail.com>
> > wrote:
> > >> >
> > >> > Please find below a proposal for a new Incubator podling named
> Apache
> > >> > RocketMQ, a fast, low latency, reliable, scalable, distributed, easy
> > to
> > >> use
> > >> > message-oriented middleware, especially for processing large amounts
> > of
> > >> > streaming data.
> > >> >
> > >> > The draft proposal can be found in the wiki at the following URL:
> > >> >
> > >> > https://wiki.apache.org/incubator/RocketMQProposal
> > >> >
> > >> > Below, please find the text for the proposal below.
> > >> >
> > >> > Thanks,
> > >> >
> > >> > Bruce
> > >>
> > >>
> > >> ---------------------------------------------------------------------
> > >> To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
> > >> For additional commands, e-mail: general-help@incubator.apache.org
> > >>
> > >>
> >
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
> > For additional commands, e-mail: general-help@incubator.apache.org
> >
> >
>



-- 
Nothing is impossible

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message