10^15 WD Red Pro

[ALFA]

are the new Western Digital WD4001FFSX the only or Red Pro the only one with the non-recoverable read error rate better than the others at < 10^15 bits?

for what I read the disks with the < 10^14 bits the 100% considered to be sure to find an non-recoverable read error rate if for 10 TB (0.01 PB), so for instance you can't have a 4TB - 10 disk raidZ-2 because it would be totally useless, even a raidZ-3

for example, if 10 disk of 4TB it would be a total of 40 TB raw storage (29 of usable storage) so when the 10 TB mark is read is almost a 100% chance of encountering a hard disk error, the failed cause the raid to rebuild and we can then expect another disk failure because all of the remaining 9 drives will have to be need read (36 TB) ,hitting another 10 TB of data, and another hard disk error, now for the RaidZ-2 there is non protecction (2 failure disk), rebuilding and reading the another 8 disk (32 TB) it sure will hit again the 10 TB mark, therefore the second rebuild is very likely to fail.

I don't remember the exact page, but the title was SAS vs. SATA from enterprisestorageforum.com

¿It's all this true?
¿What do you think guys from this new WD Red Pro HDD?

I am not an expert so I seek advice from more experienced in this forum

Thanks and kind regards to all

EDIT. found the page SAS VS SATA

 

[cyberjock]

This is statistics and engineering notations for things, so you do need to keep things on context. Yes, if you have a 10TB pool the statistics show that you are almost 100% likely to hit a URE. But, those engineering numbers are nothing more than a bunch of secret magic and they are also "rated" values. That is, your drive might be 10^30 or 10^16. As long as the drive isn't having UREs at a rate that is higher than 10^15 then everything is considered 'acceptable'.

To be honest, those numbers don't mean what most people think they do, and if you've done your homework and gone with RAIDZ2 or RAIDZ3 then I consider the pool to already be redundant against URE factors. The odds of two disks in a RAIDZ2 vdev having a URE at the same "stripe" of data in ZFS is extraordinarily small. Think "not even remotely possible in your lifetime" type of small.

 

[jgreco]

Wrong way to calculate it, Princess Cyberjock. You need to account for one of the failures being a total drive failure. This raises the probability back up to "merely unlikely."

 

[ALFA]

This is statistics and engineering notations for things, so you do need to keep things on context. Yes, if you have a 10TB pool the statistics show that you are almost 100% likely to hit a URE. But, those engineering numbers are nothing more than a bunch of secret magic and they are also "rated" values. That is, your drive might be 10^30 or 10^16. As long as the drive isn't having UREs at a rate that is higher than 10^15 then everything is considered 'acceptable'.

To be honest, those numbers don't mean what most people think they do, and if you've done your homework and gone with RAIDZ2 or RAIDZ3 then I consider the pool to already be redundant against URE factors. The odds of two disks in a RAIDZ2 vdev having a URE at the same "stripe" of data in ZFS is extraordinarily small. Think "not even remotely possible in your lifetime" type of small.

Wrong way to calculate it, Princess Cyberjock. You need to account for one of the failures being a total drive failure. This raises the probability back up to "merely unlikely."

So, i have nothing to fear if I'll go with a 10^14 10 disk 4 TB RaidZ-2?

and sorry for ask cyberjock, but what happened with you, have you had a change of heart or lost some bet? it is difficult to take you seriously with that change (just kidding, Im know you are very skilled and professional in your area) Im a long Lurker for this forums and know too you are the Dog guard admin, hope punishment be short and jgreco has nothing to do with it ;)

 

[jgreco]

So, i have nothing to fear if I'll go with a 10^14 10 disk 4 TB RaidZ-2?

I think you don't have very much to fear. It is certainly possible to lose an entire disk and then have an URE on another disk which means that the corresponding data is no longer redundant but still available. The question you need to ask yourself is "how likely is it that there will be a problem during the resilver," maybe knocking another drive offline because you're stressing it. Even at that point, your status is that you've lost redundancy entirely and any URE's on the remaining disk represent irretrievable data.

With that much data, the extra safety margin of a RAIDZ3 and a warm spare disk make me feel good. But I'm admittedly paranoid. My observation is that if you're prepared for failure then fate goes knocking on someone else's pool and you rarely even experience a single disk failure, but the moment you get daring and deploy a RAIDZ1 is the day two drives fail on you.

and sorry for ask cyberjock, but what happened with you, have you had a change of heart or lost some bet? it is difficult to take you seriously with that change (just kidding, Im know you are very skilled and professional in your area) Im a long Lurker for this forums and know too you are the Dog guard admin, hope punishment be short and jgreco has nothing to do with it ;)

I'd say it's a little friendly hazing, and anything I had to do with it is very indirect at best.

 

[cyberjock]

The iX CTO is playing with me. :)

Mysteriously I don't have rights to mess with his account? It must be a coincidence. ;)

 

[jgreco]

I noticed I got de-staffed at some point too. So I guess it's nice at least they're paying more attention to the forums.

 

[Ericloewe]

The new guys will certainly be confused when they go looking for Cyberjock's guide and the only thing they find is PersonablePrincessPeach's guide.

 

[jgreco]

"PersonablePrincessPeach" is totally impossible to remember. I can remember the princess bit but all I think of for the rest is angry sharp pointy teeth.

More like

http://i1.cpcache.com/product_zoom/...r=White&height=460&width=460&padToSquare=true

 

[DJ9]

:D

 

[Ericloewe]

"PersonablePrincessPeach" is totally impossible to remember. I can remember the princess bit but all I think of for the rest is angry sharp pointy teeth.

More like

http://i1.cpcache.com/product_zoom/...r=White&height=460&width=460&padToSquare=true

I vote for the vicious beast of Caerbannog to be Cyberjock's PersonablePrincessPeach's new image.

 

[Robert Smith]

Read the specs carefully: how many errors are allowed in that 10^15?

10 in 10^15 is the same as 1 in 10^14

 

[ALFA]

Read the specs carefully: how many errors are allowed in that 10^15?

sorry if not quite understand your question but from what I understand 10^15 disks improve the hard error rate by a factor of 10, so If you read 111 TB (0.11 PB) of data your probability of encountering a read error approaches at 100%

and upvote from the vicious beast of Caerbannog ;)

 

[Robert Smith]

sorry if not quite understand your question but from what I understand 10^15 disks improve the hard error rate by a factor of 10, so If you read 111 TB (0.11 PB) of data your probability of encountering a read error approaches at 100%

I see that Western Digital’s marketing geniuses idea worked brilliantly. Bravo, WD!

 

[ALFA]

I see that Western Digital’s marketing geniuses idea worked brilliantly. Bravo, WD!

Ohhh, I get it, Cyberjock ... I mean PersonablePrincessPeach said the same, it could be 10^30 but still you could get at URE like 10^14
well the idea its to learn, don't you think? I know WD has a very tricky propaganda.

 

[Ericloewe]

Read the specs carefully: how many errors are allowed in that 10^15?

10 in 10^15 is the same as 1 in 10^14

That's underhanded even by HDD manufacturer standards.

I knew I'd read they were still 10^14 somewhere...