You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@hama.apache.org by "Edward J. Yoon" <ed...@apache.org> on 2012/05/03 13:09:22 UTC

Hama Talk at Daum, Jeju Island

Yesterday I talked about our project @ Daum communication (very large
web portal service company in S.Korea), and just wanted to share few
pics with you -
http://blog.udanax.org/2012/05/apache-hama-talk-at-daum-jeju-island.html

They were more interested in hearing about real-time processing for
*Quick interactive service* than batch processing. And it's true that
real-time processing and tape backup of raw data is more cheaper than
maintenance of HDFS with 3 replicas on the Disk.

-- 
Best Regards, Edward J. Yoon
@eddieyoon

Re: Hama Talk at Daum, Jeju Island

Posted by "Edward J. Yoon" <ed...@apache.org>.
They will try to contribute new something related to it soon.

> Oh really? That is interesting.

Yeah, large web service creates a lot of 4GB log files in seconds.

On Thu, May 3, 2012 at 8:23 PM, Thomas Jungblut
<th...@googlemail.com> wrote:
> Wow great, didn't know I had so much impact with my blog post.
> We have to build a lot more convenience and stuff arround BSP to make it
> better usable for real-time.
>
> And it's true that
>> real-time processing and tape backup of raw data is more cheaper than
>> maintenance of HDFS with 3 replicas on the Disk.
>>
>
> Oh really? That is interesting.
>
> 2012/5/3 Edward J. Yoon <ed...@apache.org>
>
>> Yesterday I talked about our project @ Daum communication (very large
>> web portal service company in S.Korea), and just wanted to share few
>> pics with you -
>> http://blog.udanax.org/2012/05/apache-hama-talk-at-daum-jeju-island.html
>>
>> They were more interested in hearing about real-time processing for
>> *Quick interactive service* than batch processing. And it's true that
>> real-time processing and tape backup of raw data is more cheaper than
>> maintenance of HDFS with 3 replicas on the Disk.
>>
>> --
>> Best Regards, Edward J. Yoon
>> @eddieyoon
>>
>
>
>
> --
> Thomas Jungblut
> Berlin <th...@gmail.com>



-- 
Best Regards, Edward J. Yoon
@eddieyoon

Re: Hama Talk at Daum, Jeju Island

Posted by Thomas Jungblut <th...@googlemail.com>.
Wow great, didn't know I had so much impact with my blog post.
We have to build a lot more convenience and stuff arround BSP to make it
better usable for real-time.

And it's true that
> real-time processing and tape backup of raw data is more cheaper than
> maintenance of HDFS with 3 replicas on the Disk.
>

Oh really? That is interesting.

2012/5/3 Edward J. Yoon <ed...@apache.org>

> Yesterday I talked about our project @ Daum communication (very large
> web portal service company in S.Korea), and just wanted to share few
> pics with you -
> http://blog.udanax.org/2012/05/apache-hama-talk-at-daum-jeju-island.html
>
> They were more interested in hearing about real-time processing for
> *Quick interactive service* than batch processing. And it's true that
> real-time processing and tape backup of raw data is more cheaper than
> maintenance of HDFS with 3 replicas on the Disk.
>
> --
> Best Regards, Edward J. Yoon
> @eddieyoon
>



-- 
Thomas Jungblut
Berlin <th...@gmail.com>

Re: Hama Talk at Daum, Jeju Island

Posted by "Edward J. Yoon" <ed...@apache.org>.
+1

On Thu, May 3, 2012 at 8:26 PM, Tommaso Teofili
<to...@gmail.com> wrote:
> very nice !
> It'd be good to provide some more detailed documentation/examples about how
> to use Apache Hama for real time processing.
> Tommaso
>
> 2012/5/3 Edward J. Yoon <ed...@apache.org>
>
>> Yesterday I talked about our project @ Daum communication (very large
>> web portal service company in S.Korea), and just wanted to share few
>> pics with you -
>> http://blog.udanax.org/2012/05/apache-hama-talk-at-daum-jeju-island.html
>>
>> They were more interested in hearing about real-time processing for
>> *Quick interactive service* than batch processing. And it's true that
>> real-time processing and tape backup of raw data is more cheaper than
>> maintenance of HDFS with 3 replicas on the Disk.
>>
>> --
>> Best Regards, Edward J. Yoon
>> @eddieyoon
>>



-- 
Best Regards, Edward J. Yoon
@eddieyoon

Re: Hama Talk at Daum, Jeju Island

Posted by Tommaso Teofili <to...@gmail.com>.
very nice !
It'd be good to provide some more detailed documentation/examples about how
to use Apache Hama for real time processing.
Tommaso

2012/5/3 Edward J. Yoon <ed...@apache.org>

> Yesterday I talked about our project @ Daum communication (very large
> web portal service company in S.Korea), and just wanted to share few
> pics with you -
> http://blog.udanax.org/2012/05/apache-hama-talk-at-daum-jeju-island.html
>
> They were more interested in hearing about real-time processing for
> *Quick interactive service* than batch processing. And it's true that
> real-time processing and tape backup of raw data is more cheaper than
> maintenance of HDFS with 3 replicas on the Disk.
>
> --
> Best Regards, Edward J. Yoon
> @eddieyoon
>

Re: Hama Talk at Daum, Jeju Island

Posted by "Edward J. Yoon" <ed...@apache.org>.
+1

Sent from my iPad

On May 5, 2012, at 1:42 AM, Praveen Sripati <pr...@gmail.com> wrote:

> Should I open a JIRA, so that we can have a discussion and a solution
> around this?

Re: Hama Talk at Daum, Jeju Island

Posted by Praveen Sripati <pr...@gmail.com>.
Real-time processing with Hama is good, but unless it can scale up/down
automatically not sure how many would really buy it. It might be Hama or
the BSP limitation itself, but finally it's that Hama can't scale once the
job has been started. The associated costs are fixed over time.

Should I open a JIRA, so that we can have a discussion and a solution
around this?

Praveen

On Thu, May 3, 2012 at 4:39 PM, Edward J. Yoon <ed...@apache.org>
wrote:
> Yesterday I talked about our project @ Daum communication (very large
> web portal service company in S.Korea), and just wanted to share few
> pics with you -
> http://blog.udanax.org/2012/05/apache-hama-talk-at-daum-jeju-island.html
>
> They were more interested in hearing about real-time processing for
> *Quick interactive service* than batch processing. And it's true that
> real-time processing and tape backup of raw data is more cheaper than
> maintenance of HDFS with 3 replicas on the Disk.
>
> --
> Best Regards, Edward J. Yoon
> @eddieyoon