You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@hama.apache.org by Praveen Sripati <pr...@gmail.com> on 2012/04/26 17:55:12 UTC
Pros of Hama over Giraph over Graph Processing
What advantages does Hama offer over Giraph for graph processing?
After watching the following video (http://vimeo.com/40737998) both Hama
and Giraph seems to have somewhat similar API approach for graph
processing. Both of them extend/implement a vertex class and implement the
compute method.
We can may be covert this discussion into a Wiki.
Thanks,
Praveen
Re: Pros of Hama over Giraph over Graph Processing
Posted by Praveen Sripati <pr...@gmail.com>.
Suraj,
Outside review would be unbiased and that is one of the option. But,
instead of waiting for a good outside samaritan who is interested in
Hama/Giraph/Mahout and others to do an unbiased review, it would good to
get started with something. Also, not sure if someone from outside would do
it without any $.
Praveen
On Sat, May 26, 2012 at 8:34 PM, Suraj Menon <su...@apache.org> wrote:
> @Billy, completely agree, I gave a similar message at Apache BarCamp
> DC last week.
>
> @Praveen, I am sorry, I differ with you here. :) The Hama wiki should
> only explain Hama performance. In my personal opinion, this should be
> done by any interested user/developer outside. Hopefully, we would be
> proud to share that link.
>
> -Suraj
>
> On Sat, May 26, 2012 at 9:23 AM, Billy Newport <bi...@billynewport.com>
> wrote:
> > I think the difference may be that Hama provides a bsp base with graph
> > also. Giraph seems to just do graph. With Hama becoming a top level
> > apache project now, it's hard to see giraph making it also given the
> > overlap. My vote would be for giraph to combine forces and enhance
> > Hama but we will see, I guess.
> >
> > Sent from my iPad
> >
> > On May 26, 2012, at 9:10 AM, Praveen Sripati <pr...@gmail.com>
> wrote:
> >
> >> If I see the benchmarks [1], it's always Hama against Hama. While this
> is
> >> absolutely necessary to make sure Hama doesn't regress, it's even more
> >> necessary to compare (functionality, performance, algorithms etc) Hama
> with
> >> others.
> >>
> >> Couldn't agree more with Suraj on benchmarks to benchmarks comparison.
> >> Since everything is available as open source it should be possible to
> do a
> >> comparison with the same data set with out-of-the-box software without
> any
> >> fine tuning for different algorithms.
> >>
> >> Just my 2c.
> >>
> >> Praveen
> >>
> >> [1] - http://wiki.apache.org/hama/Benchmarks
> >>
> >> On Fri, May 25, 2012 at 2:22 PM, Edward J. Yoon <edwardyoon@apache.org
> >wrote:
> >>
> >>> and also, GraphLab, and Spark.
> >>>
> >>> On Fri, Apr 27, 2012 at 12:55 AM, Praveen Sripati
> >>> <pr...@gmail.com> wrote:
> >>>> What advantages does Hama offer over Giraph for graph processing?
> >>>>
> >>>> After watching the following video (http://vimeo.com/40737998) both
> Hama
> >>>> and Giraph seems to have somewhat similar API approach for graph
> >>>> processing. Both of them extend/implement a vertex class and implement
> >>> the
> >>>> compute method.
> >>>>
> >>>> We can may be covert this discussion into a Wiki.
> >>>>
> >>>> Thanks,
> >>>> Praveen
> >>>
> >>>
> >>>
> >>> --
> >>> Best Regards, Edward J. Yoon
> >>> @eddieyoon
> >>>
>
Re: Pros of Hama over Giraph over Graph Processing
Posted by Suraj Menon <su...@apache.org>.
@Billy, completely agree, I gave a similar message at Apache BarCamp
DC last week.
@Praveen, I am sorry, I differ with you here. :) The Hama wiki should
only explain Hama performance. In my personal opinion, this should be
done by any interested user/developer outside. Hopefully, we would be
proud to share that link.
-Suraj
On Sat, May 26, 2012 at 9:23 AM, Billy Newport <bi...@billynewport.com> wrote:
> I think the difference may be that Hama provides a bsp base with graph
> also. Giraph seems to just do graph. With Hama becoming a top level
> apache project now, it's hard to see giraph making it also given the
> overlap. My vote would be for giraph to combine forces and enhance
> Hama but we will see, I guess.
>
> Sent from my iPad
>
> On May 26, 2012, at 9:10 AM, Praveen Sripati <pr...@gmail.com> wrote:
>
>> If I see the benchmarks [1], it's always Hama against Hama. While this is
>> absolutely necessary to make sure Hama doesn't regress, it's even more
>> necessary to compare (functionality, performance, algorithms etc) Hama with
>> others.
>>
>> Couldn't agree more with Suraj on benchmarks to benchmarks comparison.
>> Since everything is available as open source it should be possible to do a
>> comparison with the same data set with out-of-the-box software without any
>> fine tuning for different algorithms.
>>
>> Just my 2c.
>>
>> Praveen
>>
>> [1] - http://wiki.apache.org/hama/Benchmarks
>>
>> On Fri, May 25, 2012 at 2:22 PM, Edward J. Yoon <ed...@apache.org>wrote:
>>
>>> and also, GraphLab, and Spark.
>>>
>>> On Fri, Apr 27, 2012 at 12:55 AM, Praveen Sripati
>>> <pr...@gmail.com> wrote:
>>>> What advantages does Hama offer over Giraph for graph processing?
>>>>
>>>> After watching the following video (http://vimeo.com/40737998) both Hama
>>>> and Giraph seems to have somewhat similar API approach for graph
>>>> processing. Both of them extend/implement a vertex class and implement
>>> the
>>>> compute method.
>>>>
>>>> We can may be covert this discussion into a Wiki.
>>>>
>>>> Thanks,
>>>> Praveen
>>>
>>>
>>>
>>> --
>>> Best Regards, Edward J. Yoon
>>> @eddieyoon
>>>
Re: Pros of Hama over Giraph over Graph Processing
Posted by Billy Newport <bi...@billynewport.com>.
I think the difference may be that Hama provides a bsp base with graph
also. Giraph seems to just do graph. With Hama becoming a top level
apache project now, it's hard to see giraph making it also given the
overlap. My vote would be for giraph to combine forces and enhance
Hama but we will see, I guess.
Sent from my iPad
On May 26, 2012, at 9:10 AM, Praveen Sripati <pr...@gmail.com> wrote:
> If I see the benchmarks [1], it's always Hama against Hama. While this is
> absolutely necessary to make sure Hama doesn't regress, it's even more
> necessary to compare (functionality, performance, algorithms etc) Hama with
> others.
>
> Couldn't agree more with Suraj on benchmarks to benchmarks comparison.
> Since everything is available as open source it should be possible to do a
> comparison with the same data set with out-of-the-box software without any
> fine tuning for different algorithms.
>
> Just my 2c.
>
> Praveen
>
> [1] - http://wiki.apache.org/hama/Benchmarks
>
> On Fri, May 25, 2012 at 2:22 PM, Edward J. Yoon <ed...@apache.org>wrote:
>
>> and also, GraphLab, and Spark.
>>
>> On Fri, Apr 27, 2012 at 12:55 AM, Praveen Sripati
>> <pr...@gmail.com> wrote:
>>> What advantages does Hama offer over Giraph for graph processing?
>>>
>>> After watching the following video (http://vimeo.com/40737998) both Hama
>>> and Giraph seems to have somewhat similar API approach for graph
>>> processing. Both of them extend/implement a vertex class and implement
>> the
>>> compute method.
>>>
>>> We can may be covert this discussion into a Wiki.
>>>
>>> Thanks,
>>> Praveen
>>
>>
>>
>> --
>> Best Regards, Edward J. Yoon
>> @eddieyoon
>>
Re: Pros of Hama over Giraph over Graph Processing
Posted by Praveen Sripati <pr...@gmail.com>.
If I see the benchmarks [1], it's always Hama against Hama. While this is
absolutely necessary to make sure Hama doesn't regress, it's even more
necessary to compare (functionality, performance, algorithms etc) Hama with
others.
Couldn't agree more with Suraj on benchmarks to benchmarks comparison.
Since everything is available as open source it should be possible to do a
comparison with the same data set with out-of-the-box software without any
fine tuning for different algorithms.
Just my 2c.
Praveen
[1] - http://wiki.apache.org/hama/Benchmarks
On Fri, May 25, 2012 at 2:22 PM, Edward J. Yoon <ed...@apache.org>wrote:
> and also, GraphLab, and Spark.
>
> On Fri, Apr 27, 2012 at 12:55 AM, Praveen Sripati
> <pr...@gmail.com> wrote:
> > What advantages does Hama offer over Giraph for graph processing?
> >
> > After watching the following video (http://vimeo.com/40737998) both Hama
> > and Giraph seems to have somewhat similar API approach for graph
> > processing. Both of them extend/implement a vertex class and implement
> the
> > compute method.
> >
> > We can may be covert this discussion into a Wiki.
> >
> > Thanks,
> > Praveen
>
>
>
> --
> Best Regards, Edward J. Yoon
> @eddieyoon
>
Re: Pros of Hama over Giraph over Graph Processing
Posted by "Edward J. Yoon" <ed...@apache.org>.
and also, GraphLab, and Spark.
On Fri, Apr 27, 2012 at 12:55 AM, Praveen Sripati
<pr...@gmail.com> wrote:
> What advantages does Hama offer over Giraph for graph processing?
>
> After watching the following video (http://vimeo.com/40737998) both Hama
> and Giraph seems to have somewhat similar API approach for graph
> processing. Both of them extend/implement a vertex class and implement the
> compute method.
>
> We can may be covert this discussion into a Wiki.
>
> Thanks,
> Praveen
--
Best Regards, Edward J. Yoon
@eddieyoon