My signature gives it away that my unRaid box is on the small side when compared to the monoliths that some create. In fact I'm in no real need for the extra HDDs, but since I know I'm about to add a new share to store a fair chunk of data  (and share levels always confuse me!  ) it may be time to add a 3rd drive.

 

But then I got thinking about the best way to proceed. Should I just buy another 2TB drive or look for something smaller? Buying many smaller drives does have economies of scale - you only buy the space you need when you need it. But then many drives = many mechanical devices that could fail.

 

OTOH....a 2TB drive failing (especially if a user can't afford to buy a replacement for a month or so) could be quite a headache. Plus you spread out the workload over many smaller drives, meaning each drive could last longer.

 

A subject worthy of debate? Or just a dumb question? You decide.... 

Buy 4TB drives.

There is no correlation between the 'amount' of work that a hard drive does and failure rate.  Failure rates are determined mostly by power-on hours and temperature.  Once the "teething problems" of a new generation of larger drives are resolved, they have about the same failure rate as older (and smaller) hard drives. 

 

You might not think so but the probability of a hard drive failure (in a server) is directly proportional to number of  hard drives that the system contains.  Twice the number of hard drives, twice the risk of a failure...

 

An another factor to consider is that it will cost more per TB to use smaller drives than larger ones for most unRAID users.  The only time when a smaller size hard drive setup makes sense is when you have a small finite storage requirement that will never increase.  As an example, you know that you will never, ever require more than 3TB of storage, you might built a system with four 1TB drives.   

Appreciate the posts (and apologies for the delay in responding)

 

I would have thought that the MTBF figures for hard drives would mean that the more often a drive is used, the greater the chance of it failing. It's a big leap but smaller drive = less access = less chance to fail.

 

At the same time, 10 x 200GB drives mean that there's 10 more times for a mechanical failure when compared to a single 2TB drive. So does that mean there's a balance between size of drive used and number of drive? Or is this just hypothetical drivel?? 

 

At the moment I'm just going for the "get the biggest drive you can afford" approach, whilst remembering that £5-10 more usually adds 500GB of capacity.

A few comments ...

 

=>  Modern drives have non-recoverable error rate specifications on the order of 1 in 10^14 bits.    This may seem to indicate you'll get more errors with large drives [For example, a 2TB drive contains 1.6 x 10^13 bits -- so on average there will be an unrecoverable error once every 6 reads !!    But remember that while smaller drives with the same error specifications would have far more reads between errors, it's also true that to access the same amount of data you have to do a lot more reads if you're spreading the data across smaller drives.    Note also that some older drives have 1 in 10^13 specs, so the potential for errors is likely even higher.

 

=>  The real issue is how reliable the drive is -- and the reality is that modern high capacity drives are every bit as reliable as the older units of the same quality level [i.e. desktop drive vs. desktop drive;  enterprise drive vs. enterprise drive]

 

=>  For the case you've asked above, a FAR more relevant/important issue with regard to reliability is simply adding a parity drive.    Based on your listed system specs, you've got two data drives -- but no parity.    If you need more space, it's time to BOTH add a 2TB parity drive and add another 2TB drive.    Then you'll not only have the additional space you need;  but you'll be protected from any drive failures.    Of course, you'll also have to buy a Plus license to go beyond 3 drives -- but this will also provide you additional room to grow in the future.

 

Just get the biggest drives with the longest warranties and most reliable rep.

 

a 4TB Hitachi with a 3 year warranty makes sense to me. The price per GB is not significantly higher than a 3TB drive.

 

I have a couple of 2008 1.5TB Segates that have always been a bit dodgy, but they have warranties going out to 2013. By the time the warranty is done, you will have got your money's worth out of the drive.

The whole thing of MTBF got me thinking - I looked at the WD drives (2TB) to compare/contrast:

 

- Enterprise drive (WD RE): 1.4m MTBF, 5yr warranty, 600k load/unload cycles. source

- WD Red ("designed for NAS"): 1m MTBF, 3yr warranty, 600k load/unload cycles. source

- WD Green (desktop drive): unknown MTBF, 2yr warrantry, 300k load/unload cycles. source

 

Now I've got 2 x WD Green drives (1 of them is parity so I'm covered on that front). So I think I need to consider adding a new parity drive rather then just add more storage. In terms of space usage it's not an immediate requirement - I'm only now getting close to filling 1TB worth of space. But the parity drive (I would imagine) gets used more than a typical data drive (or is that a fallacy?) - therefore it's more prudent to use a drive with the widest margin of error. Or is that yet more drivel? 

 

As a footnote the WD Red drives are not mentioned on the unRaid wiki at first glance - http://lime-technology.com/wiki/index.php/Hardware_Compatibility#Hard_Drives - is this something that needs reviewing?

 

 

Are you downloading and writing to your array all the time?

 

My usage pattern is as follows on Server1:

For about 8-36 hours straight about every 3-6 months I move files to my array.  Plus once a month I do a non-correcting parity check.  Other than that my parity drive is idle and not used I have an old Hitachi 3TB 7200rpm that I got before they were sold.  At least one of my data drives is READ every day for media playback. 

 

My usage pattern is as follows on Server2:

The second server has a similar usage pattern to Server1 but also I write backups to it every day.  I will be getting a drive with a longer warranty for the parity drive but I've been waiting on sales on WD 3TB black drives.  I currently have a 1 year warranty drive as parity on it.

 

So my recommendation is to look at your expected usage pattern and plan according.

Good point. My use is primarily for AV storage - so "write once, read many". Whilst I'm also storing documents and others things on there....again, that's more of a local backup solution and the data requirements are far smaller.

 

So there's a good rule of thumb for users that are planning their first builds - or at least the outline of one! Consider how you will be accessing the server; how often will parity be changing? How often will the data be accessed?

 

Has anyone tested these WD Red drives in UnRaid?