Best Solution for Creating a 5-Drive, RAID 5 File Server.

Factual Questions
1

As I noted in a previous post, I’m unsuccessfully trying to setup a five-drive, RAID 5 file server using a Sans Digital MobleRIAD. Since the Mac Mini I’m connecting it to is limited to either Firewire or USB, I’m unable access the full 4TB of drive space. Can anyone offer an alternative solution?

Requirements:
Must accommodate 5 (or more, but preferably not) 1TB drives
Must Use RAID 5 or RAID 6
Preferably can connect to my network over ethernet.
Shouldn’t come with an enterprise-level price tag.

2

You should be able to host 2 2TB partitions on it.

3

How do I do that and keep it RAID 5? What happens to the fifth drive?

4

I thought RAID 5 just striped your data across all available physical disks? Wouldn’t you just have 2 2-TB “stripes” (not counting parity data).

5

I’m not sure which of your threads to put this (or even if it’s worth mentioning).

You said in the other thread that speed was a low priority. What about expanding your options by considering something with Gigabit connectivity? Or if speed really isn’t all that important, what about a 10/100 Ethernet device (not that that would be painfully slow, but I don’t know what you’ll be using the data for)?

That will free you up to build a very low-end machine processing-wise, throw a basic Linux OS at it and use it as a networked drive. You’ll be well under five hundred bucks or so, and your file server can have a vast amount of untapped flexibility for future ideas.

Again, I have no idea if this is appropriate or what you’re planning on doing, so forgive me if this is way off base.

6

Hi Nobody, I answered part of your question on the other thread, hopefully some of this will be useful.

The problem: You have a hard drive enclosure with 5x 1TB drives. You want to make the largest possible single drive that the operating system can use while still offering RAID protection. However, your 32 bit O/S can only see a max 2TB drive.

RAID is a compromise, it always sacrifices drive capacity for redundancy. The higher the RAID number, generally the better the protection. The exception is RAID0 - ie. zero protection but higher performance due to reading from many drives at once.

You will have to “log on” to the management screen of the drive enclosure and experiment with the available configurations to see how best to divide up the hard drives.

RAID0: Not really an option as there is no extra data security but you could use 2x 1TB (2TB total) / 2x 1TB (2TB total) and 1x 1TB (1TB total) giving you three separate non-redundant drives to use. Data is broken up into chunks and “stripe” written to the drives in turn giving performance but no extra protection.

RAID1: Disk mirroring, but you lose 50% of the available drives per array. So you could set up 2x 1TB (1TB total) / 2x 1TB (1TB total) and 1x 1TB unassigned. You end up with 2 usable drives but have “lost” three out of the set due to redundancy ie. drives without a pair. You could also have a pair of drives mirrored, ie. (2x 2x 1TB) and 1 drive spare but you would still only have a single 2TB drive to play with. Gives good redundancy but at a high price in lost overall capacity.

RAID5: Best compromise, data is broken up into chunks and an extra parity chunk is created based on the original data. It is then stripe written to the drives in turn giving both performance and redundancy. You need at least 3 drives to make this work so you could use 3x 1TB (2TB total) and mirror the remaining 2x 1TB (1TB total) giving you two drives, one striped of 2GB and one mirrored of 1 TB.

Other RAID levels play about with striping a mirror etc and can be demanding of overhead - check Wikipedia article - eg. RAID6 can tolerate two drives failing at once where RAID5 can only tolerate one failure.

Having said all this, the best thing to do would be to see if the management software allows you to create “logical” splits. To explain, you create a RAID5 set including all 5 drives - this gives you a physical array of 4TB. But - you then say: I want to make a logical division of that array into two “software” or “logical” drives of 2TB each.

So the underlying hardware is a 5-disk RAID5 array, but two 2TB logical drives are passed to the operating system for use. Driver abstraction means the O/S won’t even know it doesn’t have two distinct drives.

That’s the best way to go ending up with 2x 2TB protected drives but if that is not possible go with one stripe and one mirror giving 1x 2TB and 1x 1TB.

7

How is that an exception? :wink:

8

This may in know way be what you’re looking for, as I’m not much of a tech guy…

Have you seen the Drobo? (www.drobo.com)
I know one of the inventors (or VCs) through another board. (Not KNOW, just know that he was a part of it somehow…)

9

Interesting unit. I wonder if its “Beyond Raid” is a sophisticated implementation of LVM? (Not that I’m close to being an expert about such things.) At $500 without drives, you’ll probably want to look at other network attached storagedevices.

10

You have one 4 GB RAID 5 array using the 5 drives, and you partition it into 2x 2 TB partitions.

11

Because there’s no protection against data loss in the event of disk failure, as there is in the other RAID implementations. RAID-0 is really a misnomer, since there aren’t any redundant disks in the array at all.

12

And again I say, how is that an exception to the original statement, “the higher the number, the better the protection”? Zero is the lowest RAID number, and zero is the lowest protection.

13

I think because it’s an array of striped disks, not a Redundant array.

I don’t think this is as important as which end of the egg to open, or that there was a great deal of pedantic seriousness there. But technically, isn’t RAID0 really SAID?

14

The OPs problem is that using a USB/Firewire interface, the maximum device size (for a 32bit OS) is 2TB - this is a limitation of the USB/Firewire Mass Storage Drivers addressing.

If the Raid array itself can partition the array and present 2 logical devices over USB/Firewire (each of 2TB), this may work.

Si

15

I’m not sure this is entirely correct. I remember when hard drives were similarly limited by the OS: we simply created partitions. It went from 32 MB to 504 MB to 2 GB to 128 GB to whatever the limit is today.

16

I’m confused. I thought this thread was about starting from scratch – else why the requirements list (e.g., LAN, price)? Is the OP stuck with the device and is looking for something between it and his Mac Mini? Or does he/she have the freedom to start from scratch.

If there’s the freedom, I still say that this is one of those cases where homebuilding will go further both cost- and flexibility-wise. A very cursory look with a $500 budget (sans drives) puts together a good Linux box, even with a RAID card )if necessary to get over the fairly standard four internal SATA connectors). I’m assuming running it headlessly, though a basic KVM may be included without cracking the budget. And since the box will primarily be just serving files, virtually any currently available board/processor combination will do the trick. Go used or discontinued and the price drops dramatically without a noticeable sacrifice in as-used performance.

17

From the MS USB Mass Storage Driver FAQdated 2004

While I am sure that the modern USB-ATA bridges do not have this 48-bit addressing limitation, it appears that the 32bit drivers in OS X and XP still only implement 48-bit addresses, thus the limitation (which the more modern 64-bit drivers do not have). If the RAID unit can be configured to present multiple LUNs (sub-devices) each of 2TB, this may avoid the problem.

eSATA does not have this limitation.

Si

18

Yes, but I’m not sure that that’s relevant because it should only apply to single drives. The device under discussion is a RAID 5 device with considerable cleverness and will likely appear as a USB SCSI device.

19

The idea with a RAID system like the one in the OP is that the clever smarts and RAID details are hidden by the config panel/tools and it presents to the OS as a simple mass storage device with linear block addressing. But it appears that the real fault is with inbuilt 32-bit OS USB Storage drivers. The manufacturers themselves state that 2TB is the maximum sized volume that 32-bit OSes can see, and the way to see a 4TB volume is to use the eSATA interface on a 64-bit OS.

But the OP does not have a 64-bit OS or an eSATA interface. I certainly would be looking for a different device with an ethernet interface and CIFS. But I have no experience of the devices - I’d build an Atom based system and use LVM, but YMMV.

Si