Hadoop production-ready?

Hi,

Sorry for the nature of the question, but can anyone
estimate when Hadoop will be "stable and
production-ready"?  I know you've run it on clusters
with 600+ nodes, but you guys (hadoop-dev) know this beast
inside out, and can spend time troubleshooting and fixing
its problems, but for somebody who is unable to do that,
when do you think it will be as stable as Nutch and Lucene
are, for example?

I'm asking because I see a LOT of fast development, but
also see that things being worked on are all non-trivial and
still very much in the core.

Thanks,
Otis

Hi Otis,

I think this is difficult to answer. Just some points from
my  
perspective. I use it in two projects in production but not
the  
latest trunk.
Map reduce works great and stable in my scale (up to 20
boxes). I  
lost several times data from the dfs so this is tricky.
Watch out that you better never touch your network setup. If
it runs  
once my experience is that it runs. However making backups
from  
important data to a "normal" file system is a
good idea.

However hadoop related problems are fixed in a amazing short
time frame.


.. just my personal perspective.
Stefan


On 07.07.2006, at 09:58, Otis Gospodnetic wrote:

> Hi,
>
> Sorry for the nature of the question, but can anyone
estimate when  
> Hadoop will be "stable and
production-ready"?  I know you've run it  
> on clusters with 600+ nodes, but you guys (hadoop-dev)
know this  
> beast inside out, and can spend time troubleshooting
and fixing its  
> problems, but for somebody who is unable to do that,
when do you  
> think it will be as stable as Nutch and Lucene are, for
example?
>
> I'm asking because I see a LOT of fast development,
but also see  
> that things being worked on are all non-trivial and
still very much  
> in the core.
>
> Thanks,
> Otis
>
>
>
>
>