If you are so inclined, you could write your own blobstore using the raw
block device, e.g., /dev/sda on Linux, which would give you total
control over block management.  However, I strongly recommend looking at
the existing blobstores, such as Swift and Riak CS, which may already
solve the problem you have and offer a lot in terms of performance,
reliability, and robustness.

On Wed, May 28, 2014 at 11:35:59PM -0300, felipe gutierrez wrote:
> ok. Thanks. It is I would like to know.
> 
> I wonder to develope a method to check if some block has the index files of
> my driver, so I could open a driver with many files only with few blocks,
> and fill the other blocks demand.
> 
> Would you point some code I could develop this?
> This is for personal stuff, but do you think it will contribute to apache?
> 
> Thanks Andrew
> 
> 
> On Wed, May 28, 2014 at 5:08 PM, Andrew Gaul <g...@apache.org> wrote:
> 
> > On Wed, May 28, 2014 at 08:40:50AM -0300, felipe gutierrez wrote:
> > > Hi all,
> > >
> > > I am using jClouds BlobStore with filesystem, so I have blocks at my
> > > FileSystem. I woould like to know if there is a way to check which blocks
> > > contain the index of all files of my driver.
> > >
> > > For example, when I format a new driver, jclouds creates
> > > {0,1,2,...,16,2040,2047} blocks. If I add more files, more blocks will be
> > > create. But I can move the blocks to other place and still open the
> > driver
> > > with the blocks {0,1,2,...,16,2040,2047}. But when I open one of the
> > files
> > > I need the another blocks.
> > >
> > > So I would like to make a routine to preserve only the index blocks. The
> > > other blocks I will move to other place when it is necessary.
> > >
> > > Did I make myself clear?
> > > Thanks in advance,
> > > Felipe
> >
> > Sorry I do not understand.  jclouds filesystem blobstore creates files
> > on disk and does not know anything about the underlying blocks.  It uses
> > the underlying filesystem index mechanism and this is opaque to Java.
> >
> > --
> > Andrew Gaul
> > http://gaul.org/
> >

-- 
Andrew Gaul
http://gaul.org/

Reply via email to