Thread: [Kosmosfs-users] Next Planned Release?

SourceForge Headquarters 225 Broadway Suite 1600 San Diego, CA 92101 +1 (858) 454-5900

Any idea when the next release will be?  Keep up the good work!

Regards,

Eric

We try to do one release every 3 months.  The next release will be in
the next couple of weeks.  In addition to the stability fixes, the new
feature is fsck (which Lohit is working on); once that is finished,
we'll do the next release.

Sriram

On Fri, Dec 26, 2008 at 10:03 PM, Eric Holmberg <er...@om...> wrote:
> Any idea when the next release will be?  Keep up the good work!
>
> Regards,
>
> Eric
>
>
> ------------------------------------------------------------------------------
> _______________________________________________
> Kosmosfs-users mailing list
> Kos...@li...
> https://lists.sourceforge.net/lists/listinfo/kosmosfs-users
>

> We try to do one release every 3 months.  The next release 
> will be in the next couple of weeks.  In addition to the 
> stability fixes, the new feature is fsck (which Lohit is 
> working on); once that is finished, we'll do the next release.
> 

Excellent, thanks Sriram.

On a side note, what is your opinion on the best usage of KFS?  I am working
on a map-reduce framework using hypertable.  For this, KFS is perfect since
I can locate the chunkserver where the data is located and push the
calculations to it.

I also have the need for a high-reliability distributed filesystem for
archiving data.  Should I use another distributed filesystem for this, or is
KFS fault tolerant enough to be my *only* copy of this data (excluding
off-site backups)?

Thanks,

Eric

On Sat, Dec 27, 2008 at 9:35 PM, Eric Holmberg <er...@om...> wrote:

> On a side note, what is your opinion on the best usage of KFS?  I am working
> on a map-reduce framework using hypertable.  For this, KFS is perfect since
> I can locate the chunkserver where the data is located and push the
> calculations to it.

Eric, sounds like a duplication of work. Maybe we could merge efforts,
as I am working on mapreduce framework for KFS + Hypertable since
right after commiting JNI stuff for Hypertable which enabled it to use
Hadoop's MR.

What do you think?

Mateusz

Eric,

> On a side note, what is your opinion on the best usage of KFS?  I am working
> on a map-reduce framework using hypertable.  For this, KFS is perfect since
> I can locate the chunkserver where the data is located and push the
> calculations to it.

Cool...You might want to consider merging Mateusz's efforts as well.

>
> I also have the need for a high-reliability distributed filesystem for
> archiving data.  Should I use another distributed filesystem for this, or is
> KFS fault tolerant enough to be my *only* copy of this data (excluding
> off-site backups)?

For archival purposes, use another instance of KFS in WORM mode; this
allows you to backup files (and you can prevent operator errors from
deleting files).

KFS is fault-tolerant: in our setting, we lose about 1% of the
machines and data gets re-replicated.

Sriram