RAID Capacity

Do you have a question? Post it now! No Registration Necessary.  Now with pictures!

Threaded View
I've had a couple of NAS boxes in the past, and have been fairly happy  
with them.  The cost may seem a little excessive given the surplus of  
Pentium 4 machines lying in the streets--maybe I should take a closer  
look at FreeNAS.

At any rate, I recently purchase a 4-bay model and set it up.  An online  
RAID calculator suggested I would get 8.17 TiB capacity.  In fact, the  
NAS even predicted that when creating the volume in its storage manager.  
  However, when it was all done, I had 8.05 TiB capacity, and (of  
course) a bit less than that available.

Whence the disparity?  I asked my NAS' manufacturer, but (thusfar) they  
have only given me "it's nothing to worry about" responses.  I'm not  
worried, but I would still like to know where the "missing" 125 GiB has  
gone.  It would be nice to have the storage, but I'm really looking to  
comprehend how it works.

Can anyone enlighten me?

= = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =

You'll find my most recent response to their tech support posted below.  
  It might contain salient details:

- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

On 4/30/2013 5:09 PM, [NAS Manufacturer's Tech Support] wrote:
 > Hi [Grinder],
 > As mentioned in the previous email the capacity is normal, so
 > there isn't much speculation to be had, unless I'm misunderstanding
 > what you're asking.
 > Regards,
 > [Tech Support Guy]

Grinder responds:

In my original posting, I went through the premises and mathematics of  
why I think there should be 8.17 TiB.  Are any of those statements  
incorrect?  Here they are in a more specific form.


a) Standard RAID-5 array with 4 identical hard drives.

b) Each hard drive has 2794.5 GiB capacity.


1) Each drive will only have 2790 GiB available capacity after the [NAS  
operating system] and swap partitions are created.

2) The capacity of a RAID-5 array with 4 identical drives will be (4 -  
1) x (available capacity).


For these premises, my understanding is that the array's capacity will  
be 3 x 2790 GiB = 8370 GiB, or 8.17 TiB.

Is that correct?  If not, what premise is bad?

This expectation, the 8.17 TiB, matches with RAID calculators, as well  
as the [NAS operating system's] storage manager in its initial calculations.

- - - - - - - -

Please note that I am not asking these questions with an expectation of  
finding a defect.  I want to understand why the straight-line  
calculations do not match with the actual result.  Surely that extra 125  
GiB has gone somewhere, and I want to understand where.

So, responding that 8.05 TiB is normal, does not really answer the  
question so much as it dismisses it.

Thanks for the time you have already spent in addressing my questions.

= = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =

Re: RAID Capacity

Grinder wrote:
Quoted text here. Click to load it


If I was figuring it out, I'd have to "do it the hard way". (Pull drive
from NAS, analyze it while connected to a non-RAID port on my desktop
computer. Perhaps make a four drive spanning volume, so that any file
system header, is present sequentially on the first drive.)

I don't have any utilities here, that "tell me everything I need to
know" about a disk drive and its setup. Just a patchwork of
half-utilities, with a lot of gaps in between.

When I needed to know where RAID metadata was stored on a pair of
drives here, I used "dd" to write zeros to both drives. So all sectors
had a known data pattern. Then, did array initialization. Then scanned
for non-zero areas later, to figure out the precise location. There
was a 5MB reserved area near the end, with a 64KB metadata block.
And in that block, sixteen occurrences of a repeating 4KB pattern,
implying up to sixteen arrays could be defined. So the forensic
technique there, starts with cleaning the disk, so you can detect

It's possible such a large array could use GPT, and the MBR
might be a "protective MBR". And thus, the MBR doesn't have
definitive data in it. And you'd need to look at the GPT
Wikipedia article, to parse the beginning of the array.
So there could be a few surprises along the way.

The ratio of 8.17 to 8.05, doesn't ring any bells. It doesn't
appear to be a TiB to TB conversion factor for example. And on
occasions where I've gone through the same kind of analysis you
have, I too have been just a touch dismayed at the size of the
space lossage. I don't expect cylinders or fractional cylinders,
would be so large as to correspond to the size mismatch.


Re: RAID Capacity

On 5/2/2013 12:33 AM, Paul wrote:
Quoted text here. Click to load it

This is not a very granular analysis, but the partitions on each of the  
4 drives look like this:

2432 MiB    ext4
2048 MiB    swap
  128 MiB    <unallocated>
2790 GiB    unknown

So, presumably the "evaporation" is in that last, big "raid" partition  
on each drive.  (Approximately 42 GiB worth)

I don't have any additional downtime to donate to the project, but  
eventually I would like to hook these same drives up to a more  
conventional RAID setup and see if it reports more capacity.

Re: RAID Capacity

Quoted text here. Click to load it

As it happens, those numbers are close to the same ratio (0.985312) I
have here on a Linux drive, file system size vs. partition size
Partition is 920754402 sectors, df -k reports 453152384 1K-blocks total
so we have 453152384/(920754402/2) = 0.984307.

I have exactly the same ratio for my two drive RAID-1 array too.

Site Timeline