Background:
Many of us have a lot of data and we all know how painful data loss can be so we want some insurance. This isn't so much about versioning or off-site backups, but rather where to keep your data and where to back it up to. The "obvious" solution is RAID5 or some other form of parity (RAID3, RAID4, RAID-Z, etc). It enables good storage efficiency (you only lose one drive per set) and, well, that's about all. The performance isn't stellar; on par with a single disk unless your controller costs more than a nicely equipped MacBook Pro. The price is also pretty high from $500 for a Drobo (USB) to $1700 for an RT5e (eSATA*).
The crazy idea:
Buy two multi-drive RAID0 enclosures and disks for each. Periodicly (daily, hourly, whatever) rsync between the two, or run a software RAID1 across the two if you need higher availability. This is faster than RAID5 (scales well, but not quite linearly, with the number of drives) and cheaper despite losing half your drives to redundancy. It also allows for a relatively easy upgrade path: add the new drive(s) to one enclosure, delete and recreate the array on that enclosure, then sync the data; wash, rinse, and repeat with the other side.
Consider the following 6 storage needs:
1TB today with low** growth
1TB today with high growth
1.6TB today with low growth
1.6TB today with high growth
2.5TB today with low growth
4TB today with low growth
If you need more than 5TB online, or are above 3TB and growing >50%/year, you're going to need something more expensive (8 to 15 drive arrays) or clumsy (just buying multiples of the options I've listed).
Prices from Newegg for the cheapest decent SATA drive:
250GB for $65,
320GB for $80,
400 and
500GB for $100,
750GB for $190, and
1000GB for $285. Enclosure prices are also from Newegg or the manufacturer:
Drobo (4 bays) for $500,
RT5e (5 bays) for $1700,
AMS DS-2340SES (4 bays) for $175, and the
AMS DS-2350S (5 bays) for $230; these prices include the cost of a PCIe eSATA controller where appropriate. Here are the results for the best value at for each of the 6 cases above, in the format of $/GB @ drives x capacity; lower is better.
1TB / low growth
Drobo: 0.80 @ 3x500
RT5e: 2.00 @ 3x500
2340SES: 0.75 @ 2x500
2350S: 0.86 @ 2x500
A single 1TB drive in a single drive eSATA enclosure is actually the winner here, at $0.66/GB.
1TB / high growth
Drobo: 1.07 @ 2x1000
RT5e: 2.00 @ 3x500
2340SES: 0.75 @ 2x500
2350S: 0.86 @ 2x500
1.6TB / low growth
Drobo: 0.60 @ 4x500
RT5e: 1.10 @ 4x500
2340SES: 0.63 @ 3x500
2350S: 0.71 @ 3x500
1.6TB / high growth
RT5e: 1.51 @ 2x750
2340SES: 0.74 @ 2x750
2350S: 0.71 @ 3x500
2.5TB / low growth
Drobo: 2250 @ 4x750
RT5e: 2250 @ 4x750
2340SES: 0.66 @ 3x750
2350S: 0.58 @ 5x500
4TB / low growth
RT5e: 0.78 @ 5x1000
2340SES: 0.66 @ 4x1000
2350S: 0.63 @ 5x750
The 2350S is also the only option to offer 3GB for high growth and 5TB for low growth, at reasonable value for either.
As you can see, the duplicated RAID0 solution is the cheapest way at most capacity points, and the RT5e is always the most expensive.
Summary:
Using a pair of quasi-realtime mirrored RAID0 arrays offers a number of advantages over a RAID5 array for external storage:
- Faster, scaling with the number of drives
- Cheaper, particularly at low capacities or with high anticipated growth
- Easier upgrade path with low risk
- Easy portability
It also has a few disadvantages:
- Twice as many external boxes
- A disk failure on the road leaves you with no ability to repair before returning to home base
* Or FW800/USB2/eSATA RT5 if you're willing to accept a 2TB volume limit
** Low growth refers to <50%/year growth, high growth refers to >50%/year growth. 50% is an approximation of annual available hard drive capacity growth and is really just used to set the mix of used/available bays. Low growth may have all the drive bays filled while high growth requires at least two open drive bays.
Top