WD Red Vs Seagate Ironwolf (NAS)

[pc-guy]

Isn't complete raid failure though related to the size of the array and the drive URE rate though? In a single drive failure mode while the array is degraded all remaining data in a raid 5 has to be read to reconstruct the data, a single URE will result in an array failure. From my reading around with consumer drives with a 10^14 URE rates arrays over 12TB in size raid 5 is worthless, raid 6 or 10 is recommended?

not sure what you are saying but I have had raid 5 with 4 drives previously and removing one drive from the raid doesn't cause raid failure. the raid works as it should. as a matter of fact it was completely hot swappable. as soon as i put a new 4th drive it the raid software just goes about doing its thing at the background rebuilding the raid again. the failure tolerance is always 1 drive in a raid 5 arrangement.

In raid 6 it may be 2 drives, but from what i read raid 6 with 2 drive failure, it takes a while and you will see significant performance degradation. plus you probably wouldn't run raid 6 in a 4 drive arrangement.

 

[Armageus]

not sure what you are saying but I have had raid 5 with 4 drives previously and removing one drive from the raid doesn't cause raid failure. the raid works as it should. as a matter of fact it was completely hot swappable. as soon as i put a new 4th drive it the raid software just goes about doing its thing at the background rebuilding the raid again. the failure tolerance is always 1 drive in a raid 5 arrangement.

The risk is that whilst rebuilding a RAID5 array, you put a lot of stress on the remaining disks, in order to read the remaining data and parity stripes. If one of the remaining drives suffers a Unrecoverable Read Error (URE), then that disk will be dropped by the controller, and the array will fail completely. With the manufacturer stated URE rate of consumer drives, the chance of that happening statistically is almost guaranteed on larger arrays.

RAID6 helps to some degree, but essentially just postpones it as a failed disk still needs to be rebuilt. RAID10 is certainly preferred for most enterprise use, as you are generally exposed to less risk during a rebuild (because it is a lot simpler/quicker to mirror a single disk)

 

[pc-guy]

i think you got a point there...but

the ironwolf is rated @ 1 in 10^15 = 1 in 1000TB reads.

I will check my knackered 3TB drive for total reads. I think it is still well covered.

 

[ToTheMax]

i think you got a point there...but

the ironwolf is rated @ 1 in 10^15 = 1 in 1000TB reads.

I will check my knackered 3TB drive for total reads. I think it is still well covered.

Most consumer drives have a URE rate of 10^14 bits or less which only equates to around 12TB. If your raid 5 array is around that size or larger, if one drive fails it'a statistically likely another will fail during the rebuild. URE rates are important to consider in raid 5 because it's against the total size of the array and not a single drive.

 

[pc-guy]

Sorry I might be dumb here, but why is a complete rebuild of the raid will meet the URE of a single drive?

Say URE 10^14 = 12.5TB

12TB raid5 each drive still only has 4TB data to be read and rebuilt - how is it a single drive will all of sudden get 12TB?

3 drives with 10^14 each is 3x10^14 = 37.5TB (yes 12/37.5 = 32% which is very high in terms of failure rate)

also going back to Ironwolf, the 6TB drives are rated as 10^15 thats 125TB (not 1000TB as I thought the URE is in bytes but turns out it is in bits)

 

[pc-guy]

http://www.raid-failure.com/raid5-failure.aspx

the method used here seems to make sense to me. whilst there is a large risk of raid rebuilt failure it is not 100%.

 

[Armageus]

Sorry I might be dumb here, but why is a complete rebuild of the raid will meet the URE of a single drive?

Say URE 10^14 = 12.5TB

12TB raid5 each drive still only has 4TB data to be read and rebuilt - how is it a single drive will all of sudden get 12TB?

Surely it's the additive probabilities of each drive.
= 4TB/12.5TB + 4TB/12.5TB + 4TB/12.5TB = 3x0.32 = 0.96 or 96% chance of one of the Drives coming across a URE

 

[pc-guy]

Surely it's the additive probabilities of each drive.
= 4TB/12.5TB + 4TB/12.5TB + 4TB/12.5TB = 3x0.32 = 0.96 or 96% chance of one of the Drives coming across a URE

I can't do stats

you may be right in that simple calculation. but i don't actually believe or think it is that simple.

The methodology and calculation I linked in earlier suggests a success rate of rebuilding 4x4TB (10^14) in raid 5 of 27.8% - very high failure rate indeed.

raid 10 it is then

 

[ToTheMax]

I can't do stats

you may be right in that simple calculation. but i don't actually believe or think it is that simple.

The methodology and calculation I linked in earlier suggests a success rate of rebuilding 4x4TB (10^14) in raid 5 of 27.8% - very high failure rate indeed.

raid 10 it is then

This is what I went with as well, it'll also be far faster than any raid 5 setup in both day to day usage and in the event of a rebuild. Raid 5 with large drives simply isn't worth it, you might as well just strip the data for all the redundancy protection it gives.

 

[ToTheMax]

URE rates are a major consideration in raid 5 setups during a rebuild because the entire contents of the array must be read in order to reconstruct the missing disk, this is why failure rate is so high. It's also really slow when considering > 10TB, could be days or weeks for a rebuild depending on what else the array is doing.

In comparison to raid 10 reconstruction is a simple single drive copy operation, so much less data is read and no complex XOR operations need to be done (raid 5). Rebuilds are fast.

 

[Yemster41]

For those of you worried about WD pricing, hopefully good news is just around the corner. We have reduced the pricing for our WD Red Pro range as of yesterday, hopefully you will start to see the new prices at OCUK soon. No news on WD Red price moves just yet but I will keep you updated.

 

[wonko]

For those of you worried about WD pricing, hopefully good news is just around the corner. We have reduced the pricing for our WD Red Pro range as of yesterday, hopefully you will start to see the new prices at OCUK soon. No news on WD Red price moves just yet but I will keep you updated.

Does this mean we have a WD rep back on the forum or is this unofficial? Either way, welcome back!
I could have done with a rep recently while having a shocking time getting a WD drive replaced under warranty.

 

[GWANGY]

Dont forget the third leg ... toshiba. I dont know whether they do any NAS drives tho. I have used wd green drives until yesterday that ahd 26000+ powered on hours , that was normal daily use , no errors or disk faults. Now using wd reds 6Tb .. have a toshiba 5tb which is fast but noisy (7200).
If its just media wot about some sort of file-by file protection (??) so that you lose only one file on error

 

[Mercutio]

The URE rating thing, same as likelihood of drive failure, it's a compound of each drives individual failure stat.

When rebuilding you'll have X drives left, all of which have to NOT have a URE for the rebuild to work. The URE chance is therefore against the whole size of the raid.
A good hardware raid card helps with this. They'll perform patrol reads to check data and detect (and correct) URE's before they bite.

 

[Skiddley]

For those of you worried about WD pricing, hopefully good news is just around the corner. We have reduced the pricing for our WD Red Pro range as of yesterday, hopefully you will start to see the new prices at OCUK soon. No news on WD Red price moves just yet but I will keep you updated.

I see OCUK are stocking Pro 10TB drives now, but no non-pros. Any news on this?