I am also agree with suggestions of Ted and Tommas. Since Meerkat will be developed on top of Hama, it can be managed under Hama project like hama-graph. In addition, developing Meerkat will not make Hama changed largely (even not at all), because Hama is generic BSP framework.
On Mon, Apr 28, 2014 at 6:02 PM, Tommaso Teofili <tommaso.teof...@gmail.com>wrote: > > > > 2014-04-28 10:45 GMT+02:00 Ted Dunning <ted.dunn...@gmail.com>: > > Edward, >> >> Sub-projects are generally frowned on. Incubator projects can graduate >> eventually to a top-level project, but starting a project as a sub is not >> so good. The incubator docs on this are pretty good reading. >> >> But if the committer community for your real-time BSP is (or can be) the >> same as the committer community for Hama itself, you might want to just >> fold this new code directly into Hama itself. No need in that case for a >> separate project. >> >> If the committer community is very different, then a separate project is >> warranted. >> >> I don't think that the Hama development is such a high bandwidth thing >> that >> splitting is required. To my mind that says that joining groups together >> is better than breaking them apart. Together, the two efforts can feed >> off >> each other. Apart, they could each run down due to lack of interest. >> > > +1 to Ted's suggestion, we can create a separate directory on trunk for > Meerkat (or a branch) and eventually have separate release processes for > Hama core stuff and Meerkat stuff if that's needed. > > >> >> You will know much better than any of us the details of your communities. >> > > as far as I can see Meerkat is being developed by Edward and another > committer so it should be relatively straightforward to let him/her join > the Hama community. > > My 2 cents, > Tommaso > > >> >> >> >> >> >> On Mon, Apr 28, 2014 at 4:23 AM, Edward J. Yoon <edwardy...@apache.org >> >wrote: >> >> > Hi guys, >> > >> > As some people already might know, I'm recently working on real-time >> > data processing project on top of Hama BSP model, called Meerkat[1] >> > (currently only few developers (from two organizations, Kakaotalk and >> > DataSayer) are involved in this project). >> > >> > According to our internal study, Storm-like DAG-style and >> > fault-tolerant streaming processing framework can be implemented on >> > top of Hama BSP model. And, furthermore, we're thinking, it may be >> > possible to connect to other BSP applications using some smart input >> > and output formats in the future e.g., streaming graph or learning BSP >> > applications. Thus, we believe that we may be able to have an >> > next-generation architecture that processes and analyzes the data >> > rapidly in real-time, beyond complex old-style data collecting, >> > storing, ordering, processing, and analyzing architecture. >> > >> > Does it make sense to you? If so, I'd like to start to move from >> > github to ASF soon, although this is very *early* stage. Because, it's >> > highly related with Apache Hama and others. Hama sub-module or >> > sub-project, (or Apache incubator?). Which is best you think? >> > >> > I'm also CC'ing general@i.a.o to see more feedbacks (from Apache big >> > data OSS communities). >> > >> > 1. https://github.com/datasayer/meerkat >> > >> > -- >> > Best Regards, Edward J. Yoon >> > >> > --------------------------------------------------------------------- >> > To unsubscribe, e-mail: general-unsubscr...@incubator.apache.org >> > For additional commands, e-mail: general-h...@incubator.apache.org >> > >> > >> > >
