[SATLUG] Backup Recommendations

[Robert Pearson]

On 8/13/05, Travis H. <solinym at gmail.com> wrote:
> I looked at this, and unfortunately it doesn't support NFS.

It supports SMB and CIFS. It is IP attached storage. You put it on
your network, configure it and then any box with access to your
network can see and use the storage. Just like if it was direct
attached to a storage server and served with NFS. This is better for
small shops because it doesn't require a dedicated machine that must
be on at all times. Yes, it is a single point of failure (SPOF) just
like the dedicated NFS server with one power supply and one power
cord. With the TeraStation, or LinkStation or KuroBox, you can have
more than one of them so the primary is mirrored by the secondary. Or
you can zone it. The TeraStation (LinkStation) is a RAID box. The
KuroBox is not.

If NFS connections are what you understand then you should go with
NFS. This means setting up a storage server with all the disks inside
that server and running NFS as Eli <eli at then7.com> recommends in his
reply. Traditional is traditional because it works and people
understand it.

> When I looked at this, it appeared to just be an external drive enclosure.

It is much more than that. Most external drive enclosures support USB
or FireWire or both. The KuroBox (KB) supports the IP network
interface as its primary interface. You can use the USB port to expand
the storage size from the one 250GB disk drive. See the Kurobox Wiki
for adding a second drive.
See the review at---
<<http://www.tomsnetworking.com/Reviews-171-ProdID-KUROBOX.php>>

Also look at the KuroBox Wiki at <<http://www.kurobox.com>>. Lots of
interesting stuff there.

For example you could put one 250GB drive in each of four KuroBoxes
(KB) on the network and any computer could access any storage. You
can, and should, restrict the access, just like NFS, to storage that
should not be shared.

The 250GB drive size limit seems to be based on the power supply capability.

> > Block, in addition to file, level replication is desirable. This might
> > have to be a Long-Term goal.
> 
> There is something called the Linux network block device.
> Might be something similar.

Block means the basic file system block size. Like 512, 1024, 2048, or
4096 bytes per block.
Files are built with these blocks. If you had a way to quickly
determine only the blocks that had changed in a file and then only
copy those blocks, replication would be much faster.

File means if the Modification time on the file has changed then the
entire file is copied.

Another approach is the modified file/block method used by "rsync". A
"diff" is run between the file to be replicated (source) and the
previous replication (destination). A "delta" file of the difference
is created and that is copied to the destination. This difference file
is then merged with the destination file to create the replicant
(copy) of the source file. "rsync" has a very good algorithm for this.
This algorithm has been developed over a long time starting with SCCS
(Source Code Control System) and RCS (Release Control System). The
SCCS/RCS concepts were carried forward by the first VCS (Version
Control System) and on through PVCS, CVS and Subversion.

> > My long-term goal is to use P2P (Peer-to-Peer) for all replications.
> 
> By that you mean.... you're going to let people download anything from
> your hard disk, and if it dies you're going to go download all your
> files back from these people?  You can't mean that.  What do you mean?

P2P is a very powerful tool. You can install P2P on your machines and
block all external machines from seeing them. You can also block
internal machines. You can define what files or Information is
available to what machines. It is in the setup. You want to be careful
of the P2P software you install. Most of the Freeware/Shareware does
not allow this level of isolation. You can write your own P2P but the
commercially available products are much more powerful. There are some
P2P backup products that look interesting. I'm waiting to see how they
survive.
I would proceed with caution today but P2P in business could take off
like wildfire.

Thanks,  Robert