Rich Braun wrote:
> ...the unRAID sales pitch smacks of a solution in search of a problem. 
> 
> The built-in Linux RAID5 does it quite well for me, allowing mismatched
> drives...
> 
> With terabyte drives in the $50 price range, I can't see a situation
> where kernel-based software RAID1 or RAID10 wouldn't be good enough
> (performance and pricewise) for virtually any demanding situation.

The most appealing aspect of unRAID is the way it handles capacity
upgrades - either adding additional drives, or increasing the capacity
of existing droves.

Further investigation into unRAID shows it falls far short of its
potential, but much of the infrastructure is there to make it possible
to fully automate capacity upgrades once the media is physically
inserted into the NAS.

In contrast, describes the steps involved in replacing one of your 1 TB
drives in a 4-drive RAID5 set with a 2 TB drive and making use of that
additional space.

Earlier in the thread we found plenty of flaws in the unRAID approach,
but it is getting closer to an idealized expandable pool of storage.
Upgrading your storage capacity really shouldn't be harder than
upgrading your RAM capacity.


> If I want some of the benefits that unRAID promises, namely the
> ability to recover an entire filesystem from a single drive, then I
> use RAID1.

A storage model you described on this list several yeas ago is about the
closest approximation to an expandable storage pool that I've seen using
stock Linux. This is where you use RAID1 sets with a filesystem that
supports expansion. Periodically you buy whatever drive is at the sweet
spot for cost-per-capacity, and you replace the smaller drive in the
set. Steps are something like:

1. mdadm command to remove smaller drive
2. partition new drive
3. mdadm command to add new drive to array.
4. resync array. (and wait...)
5. expand partition or LVM.
6. file system command to expand FS.

Aside from this being more complicated than necessary, your files are
vulnerable to hardware faults during steps 1 through 4, and subject to
software faults during steps 5 and 6.

 -Tom

-- 
Tom Metro
Venture Logic, Newton, MA, USA
"Enterprise solutions through open source."
Professional Profile: http://tmetro.venturelogic.com/
_______________________________________________
Discuss mailing list
Discuss@blu.org
http://blu.org/mailman/listinfo/discuss

Reply via email to