Travel Technology - Can someone walk me through multiple drive/RAID enclosures?

I'm tired of dealing with multiple hard disks, hubs and the like. I have seen several RAID enclosures for less than $150 that hold 4 - 5 drives (and 6, 8 12 bay ones for not a bunch more). I am also tired of dealing with occasional failing drive. At the end of the day as internal drives are generally a bit cheaper than externals, I figure that it would cost the same to load one of these up with 1.5 TB drives. The question is, what am I looking for and what am I missing?

I don't know if you are missing it, but your drive failure rate will not go down with RAID. In fact, if the MTBF of each physical drive is 4 years, and you have 4 drives, your MTBF for any 1 of the drives will be 1 year. You really should have spares around if you do this assuming you are going for mirroring or parity RAID (or whatever that's called these days.)

Anyway, I guess the point is you have to take the cost of at least 1 spare drive into account if you are planning to do it properly.

-David

Hi hfly, I'm not sure I understand your question. LIH Prem explains MTBF but I think you are referring to instances where you've had single drives fail which have brought down your OS and you want to shift to RAID to get over this problem?

If this is the case then yes RAID will help you get over this issue (Although it's not 100% fool-proof - I've seen faulty RAID controllers and MBs bring down perfectly good systems) but it's better than single drives.

The key to implementing RAID is deciding what's important to you. You mention using external enclosures which I presume are for data but you should take into account your OS too. Generally OS boot volumes should be mirrored (RAID1) as there's a lot of I/O on them and mirroring is quicker than RAID but comes with a 100% overhead in terms of drives. Data volumes should be RAIDed according to what you want to achieve from them, reliability, access speeds etc. Usually RAID5. There are many types of RAID configuration. To start with RAID5 you need at least 3 drives but with 3 you lose 33% of the total capacity. With 4 drives you lose 25%.

If you're looking to protect a basic system with a single drive then I'd say buy another and mirror the drives. Most OS these days allow you to mirror directly through software without the need to by expensive hardware. If you've a machine with a RAID controller on the motherboard then you can do it through this instead, it will be quicker.

The most important thing though is to ensure you monitor the drives and replace them when they start to go bad. Make sure you have something set up to do this. Also make sure your setup is the best it can be too in terms of power, UPS, surge protection, dust, humidity etc.

If you can clarify your situation we can help you more.

The quick skinny - RAID is for redundancy. If a drive fails, you're not sunk. Assuming you replace said failed drive before another one goes, you continue to work uninterrupted.

RAID0 - not true RAID, as there's no redundancy. When you write data, it will be spread between the drives. It will perform faster than a single drive by virtue of writing on two drives (nearly) simultaneously. The danger is that if 1 drive fails, you lose everything, as data is split between the drives.

RAID1 - mirrored drives (any multiple of 2 drives). Redundancy, but with a modest performance hit. If 1 drive fails, the other contains everything.

RAID5 - info spans multiple drives like RAID0, and parity information is written at the same time. Minimum drives: 3. The parity information requires 1 full drive's worth of space, so 3 drives give you 2 drives worth of data, 4 give you 3, etc. If 1 drive fails, the parity information allows the RAID controller to reconstruct the missing portion and will allow you to continue in a degraded state until the bad drive is replaced or resynced. Performs better than RAID1 due to spanning.

RAID10 - take a RAID5 array and mirror it (sort of). Lots of drives. Not often used.


Important considerations:

1) If you RAID0, have a decent, full-system, backup. I use RAID0 for my main computer, but it gets backed up nightly to a megadrive using Drive Image 7.0 (no longer available), and I've restored my system several times from images so I'm comfortable with the system.

2) Since we're looking at external housings, the connection interface is likely to be more important than the RAID type. USB 2.0 will seem ponderously slow when transferring tens of gigabytes of data. Gigabit Ethernet interfaces are only useful if the computer's NIC and the switch are also gigabit, and even then it will seem like molassas in winter for the multi-gigabyte transfers.

I'd be looking at E-SATA interfaces, but probably only 1 in 3 external RAID housings offer it, and you may need a card for your computer to be able to connect to it.

3) When looking at RAID housings, carefully read review portions that refer to the RAID software. Lots of them work fine for one type of RAID but seem less than stable with others. Some only simulate certain RAID formats.

Be aware that if you use anything other than RAID1 and your RAID controller fails a year or so from now, you may be SOL with any replacement other than your exact same model. Those drives may or may not transfer to another housing or controller. With RAID1, you would be fine, as each drive could be used without RAID.

RAID10 - take a RAID5 array and mirror it (sort of). Lots of drives. Not often used.

This one isn't quite right. RAID 1+0 is a stripe set of mirrored pairs and RAID 0+1 is a mirrored pair of stripe sets. They have rather different performance metrics and drive failure accommodations. RAID 15 would be a mirrored pair of RAID 5 arrays. And while 15 is rarely used 0+1 or 1+0 are growing in popularity in some circles. There is also RAID 6 with two parity bits rather than just one so that additional drive failures can be accommodated.

At the consumer level RAID 1 or 5 is what you are likely going to be dealing with.

With 1.5TB drives, you should not rely on RAID5, but should use RAID6. Lots of articles on this - basically, it takes so long to rebuild after a drive failure, that a second failure during the rebuild is likely and would result in data corruption. RAID6 effectively gives you triple redundancy.

I have no problem factoring in the cost of an extra drive or two to make it work, 1.5 TB drives are so cheap now, and 2TB's on the way, that it is not a problem.

I guess that the main question I have is what is a good solution? Is it the Drobo? which is pricier than other enclosures that I have seen, or is there another 4,5, 8, 10 or 12 bay solution that someone can suggest?

Also how much extra speace/redundancy does RAID6 use over RAID5?

I have no problem factoring in the cost of an extra drive or two to make it work, 1.5 TB drives are so cheap now, and 2TB's on the way, that it is not a problem.

I guess that the main question I have is what is a good solution? Is it the Drobo? which is pricier than other enclosures that I have seen, or is there another 4,5, 8, 10 or 12 bay solution that someone can suggest?

Also how much extra speace/redundancy does RAID6 use over RAID5?

My understanding of the drobo enclosure is that it allows different sized disks to be used in the RAID array and leverage the full capacity of each. If you are buying a bunch of disks that are all the same size to begin with I don't see much benefit there.

A RAID6 array has two parity bits, so two of the disks in the array are sacrificed in terms of capacity. In an 8 bay solution that is a single array you'd have the usable capacity of 6 drives in a RAID6 config and 7 drives in a RAID5 config.

If you would not mind having a look, how would something like this

http://store.sansdigital-shop.com/totr8baysato.html

or this be in your opinion?

http://store.sansdigital-shop.com/totr5baysato1.html

I do not think that the first one provides (useful) RAID at all. It will do stripe sets (RAID 0) but that offers zero redundancy. Also, I would be concerned about the limited bandwidth that a single USB 2.0 connection offers to the disks.

The second one seems to have some RAID options available but it seems that it is all software based. At least it is using eSATA 3Gbit spec for more throughput.

I'm not a fan of software RAID at all. If you lose the OS you've potentially lost your data. I much prefer hardware RAID, even though it is more expensive. I wish I had a model I could recommend to you but I haven't found one that I have been willing to buy yet. And I really need one; I'm all full on my home system and have a LARGE stack of DVDs I need to rip.

I'm not a fan of software RAID at all. If you lose the OS you've potentially lost your data. I much prefer hardware RAID, even though it is more expensive. I wish I had a model I could recommend to you but I haven't found one that I have been willing to buy yet. And I really need one; I'm all full on my home system and have a LARGE stack of DVDs I need to rip.

I disagree with this. I'm a big fan of software RAID if you can afford it (the performance hit that is). With a hardware RAID controller, if the controller goes down, you need an identical one to replace it. Otherwise your data is hosed and the RAID array does nothing. So you immediately have to buy 2 hardware RAID controllers, with one sitting in storage for when the first one dies.

I run software RAID with my boot OS on another, small, cheap disk. This then mounts the RAID at boot. So if my OS goes down, I can just re-install it, and my RAID array is just fine. I use Linux as my OS, but the same can be said with Windows or Mac OS. You can even replace the OS disk with another and still access your RAID array.

The only real advantage to using a hardware RAID controller is performance. With software RAID, you have to compute the parity and checksums with the CPU, taking away processing time from other applications. So if you can afford that hit, like I can, software RAID should be a primary consideration.