Dual parity plus a cold spare? Maybe I'm over thinking this...

In researching drives for my first unRAID build, my first thought was to buy three of the biggest drives I can afford - one parity and a couple for storage (plus a couple SSD's for cache, but that's not important for the purposes of this discussion).

"Wow, the 8TB WD Reds look great, kind of pricey, though I may be able to swing it...OOH! 10TB IronWolf! (drool) Probably run kind of hot, and three would be $1500 - OUCH! It would be 18TB+ to start out, though..." ...and so on.

Then I realized that bigger disks means *very* long times for preclear, rebuilds, etc. If one drive goes down, I'm looking at days just to preclear a replacement, and that's *after* I get it through warranty replacement, and *before* throwing it in to be rebuilt - all while the remaining disks remain unprotected.

"No problem, I'll run dual parity, but that means I then have *two* drives not counted in my total available storage capacity. Ugh. Not great, but maybe worth it for peace of mind. I could just go with 4x6TB drives, I guess...less than 11TB of space.  "

I guess it's just a balance that everyone must strike according to their own preferences. Of course any important documents and photos will be backed up someplace online, but I don't particularly enjoy the thought of re-ripping large numbers of my Blu-ray movies. What are everyone else's thoughts?

Is high capacity the top priority?

Are you willing to use smaller drives, if it lets you go with dual parity?

Do preclear/rebuild times with 8-10TB drives concern you at all?

I suppose I could go with dual parity and keep a precleared cold spare on hand...if I went with 2 or 3TB drives. 

My server only has 6 3.5" bays so going more small disks isn't really an option. I reuse my old disks and bought 2 8TB disks (at that time, 8TB was max available at reasonable price - I aint paying 1 large for 2 disks).

My policy is whenever I need upgrade, I would buy the largest available (so at this time would be the 10TB IronWolf) and use it as parity and swap the current parity into the array.

In term of single or dual parity, it's all about probability in my opinion.

According to backblaze, their worst HDD failure rate in 2016 is 12.57% (WD 2TB) and worst cumulative is 10.56% (again WD 2TB).

12.57% for a 6-disk array is about 3/4 of a disk. Hence for my array, single parity would be "sufficient" - for lack of better words (I do preclear all disks before adding to array, even my good old reliable proven-track-record ones).

I imagine a dual parity would be more appropriate for someone with more than 8-10 disks.

Of course dual parity is safer than single one but then why not offsite triple-backup across 3 continents etc. Disks are not free so we have to consider diminishing returns.

In term of pre-clearing time, I was a lucky one who ALMOST lost some really precious data and the waiting time is trivial in my mind.

First thing to help confuse you even further: 

    https://www.backblaze.com/blog/hard-drive-reliability-stats-q1-2016/

As you can see from this data the odds are in your favor of not having two disks fail within a short period of time.  In my opinion, If you have less than , say, eight data disks, dual parity will probably not buy you much protection.  (Your house is probably more likely to burn down!)  UNLESS you just start up the server and ignore it until it simply quits working.  (Bad part about this report is that it comes out too late to avoid buying disks like the Seagate ST4000DX000  and WDC WD60EFRX!)

Plus with unRAID, you only lose the data on the disks that fail.  So if you have six data disks with data on them and two of those fail, the other four disks are readable and usable.

My strategy is to buy the largest disk possible in the 'sweet spot' of TB/$.  I believe right now that is in the 3-4Tb range.  I would start up with (say) seven 3TB (one parity and six data disks) for a total storage of 18TB.  (That would be about the same cost as three 6TB drives.  I would also not buy them all at once as that spreads the risk of buying into a manufacturing lot with a manufacturing defect in it! That is one of BackBlaze's problems because they want their disk purchases at the lowest possible cost!)  As you fill up your array, add more data disks.  When you reach the point where you are running out of space in the server box, then look to see if the sweet spot is now a much larger disk size, You buy a new parity disk and move the old parity as the last 3TB data drive.  Now as disks fail (or you need more space), you replace existing disks with the larger drives.

Quick comment about the WD 2TB.  You will  notice that the number of disks is small.  I suspect that these are the last few disks of that generation and they are on the 'wear-out' portion of the bathtub reliability curve.  That data show what can happen to you if you never replace older disks until failure and just keep adding new ones to an array.  You are looking at the conditions for the dual disk failure. 

You might be looking for the $/TB yield.

The attached spreadsheet might help.

Blue areas are input; your target capacity needed, and the current prices for drives.

Cost-per-TB.zip

Another thing to consider is prices tend to go down, so if you don't need the capacity and you buy now instead of later you will likely pay more. This is one of the reasons I don't keep a spare. If I really need to I can just shutdown and wait on Amazon Prime.

Also, dual parity makes more sense the more data drives you have. I only have 5 so it's debatable whether my 2nd parity would be better used as data instead. I mostly did it just for the experience.

Of course, if you're the sort that tends to ignore problems (despite V6 notifications) then more protection is probably good. There have certainly been people on the forum who just set it and forget it until one day they show up with their first post with 2 failed drives.

There have certainly been people on the forum who just set it and forget it until one day they show up with their first post with 2 failed drives.

And they will simply be showing up with a post with three failed drives...   

You might be looking for the $/TB yield.

 

The attached spreadsheet might help.

Blue areas are input; your target capacity needed, and the current prices for drives.

Glad to see I'm not the only OCD-afflicted with a color-coded, formula-laden, conditionally-formatted HDD capacity spreadsheet.