Notifications
Clear all

Certify ... Effective Number Of Passes?

(@viperss)
New Member Customer

You recommend 3 passes when you certify a hard drive; is there any statistical advantage to this versus 1 pass? I ask because I just finished certifying an 8 TB hard drive via thunderbolt 2 and it took almost a week. It really seems impractical to run 3 passes especially as hard drives continue to get larger. I have 4 more 8 TB hard drives to certify and I would like to know the confidence/risk of 1 pass versus multiple passes.

For example, if you tell me that your data says 1 pass gives an 80% confidence, 2 pass is 90% and 3 pass is 95%, then I would consider the 3 pass. But if 1 pass is 95% confidence, 2 pass is 97% and 3 pass is 99% then I might just stick with 1 pass as the rest of the passes are just diminishing returns for me.

Thanks for any input.

Quote
Topic starter Posted : 14/12/2017 4:31 am
(@softraid-support)
Member Admin

BTW: You can certify all disks at the same time!

If you only do a one pass certify, then all we do is write zero's to the disk and read them back. That is not effective, as the disk starts with zero's, and the sectors are not changing the bits/bytes it starts with.

So a two pass certify is the minimum. One pass to write a pattern, then two to put the disks back to zeros. But that may not catch "weak" sectors, that need to be reallocated later.

The reason we recommend three passes is we did a lot of testing, and three passes was enough to ensure that every sector is tested to read/write reliably, and catch all the sectors which will reallocate in the near future.

Statistically, disks which reallocate even a single sector have a far higher early failure rate. You want to catch those drives BEFORE you put them into service.

Consider the difference between discovering a disk is "failing" after 9 months (when you get to that part of the drive where the bad sector(s) are), or on day one, before you even put it into service. In the latter case, you can replace the drive with a new one, whereas after a couple months, you generally get a refurbished (previously failed) drive from the manufacturer.

There are two main benefits to performing a disk certify:
Make sure every sector can be read and written to reliably. (Accurate data)
Test every sector to eliminate drives which are likely to fail prematurely. (Predicting failure)

ReplyQuote
Posted : 14/12/2017 12:06 pm
(@j-a-duke)
Active Member Customer

You recommend 3 passes when you certify a hard drive; is there any statistical advantage to this versus 1 pass? I ask because I just finished certifying an 8 TB hard drive via thunderbolt 2 and it took almost a week. It really seems impractical to run 3 passes especially as hard drives continue to get larger. I have 4 more 8 TB hard drives to certify and I would like to know the confidence/risk of 1 pass versus multiple passes.

That’s rather strange-the length of time that your certify is taking.

I usually certify multiple disks simultaneously and I’ve done 3 x 10TB (WD Red 5400 RPM) which combined only ran for ~5 days.

My certification setup is a CalDigit T3 Thunderbolt 1 and a mid-2011 Mac Mini 2Ghz i7QC. This machine also runs our Retrospect backups to a Areca ARC-8050T1 array each night. Most 8T certifications run in ~3 days.

Faster drives will certify faster. I’m testing some Toshiba 7200 NAS drives and found that they are about 20% for the same size, IIRC.

I haven’t tried using a TB3 Mac with a TB3 enclosure, but would not expect to see faster times since the process is limited by disk speed.

Cheers,
Jon

ReplyQuote
Posted : 15/12/2017 8:39 am
(@viperss)
New Member Customer

That’s rather strange-the length of time that your certify is taking.

I usually certify multiple disks simultaneously and I’ve done 3 x 10TB (WD Red 5400 RPM) which combined only ran for ~5 days.

Hi Jon,

Not sure why it's taking so long. I'm certifying 8 TB WD Red 5400 RPM in an Akitio Thunder 2 Quad but I'm not using the other 3 slots because I wanted to try one drive first and was worried that sharing the bus might slow everything down. I have a OWC thunderbolt 2 hard drive mount that I can try to see if there is any speed change.

ReplyQuote
Topic starter Posted : 17/12/2017 6:48 am
(@softraid-support)
Member Admin

Considering Thunderbolt 2 can support >1GB/s, you can easily certify 4 drives at the same time with no time impact over doing a single drive.

ReplyQuote
Posted : 17/12/2017 3:24 pm
(@dhanson)
Active Member Customer

It is just a painfully long process with the drive capacities we now have.

I’m currently certifying a handful of new Toshiba 3TB drives using an older MacBook Pro over FW800 and it’s taking about 3-4 days each. If I try doing two at once daisy chained on FW800 it’s doubling the amount of time needed.

ReplyQuote
Posted : 18/12/2017 12:58 pm
(@softraid-support)
Member Admin

FireWire is way too slow. You need a faster bus!

Over FireWire, 3-4 days each seems fast, even. Thunderbolt will allow a drive to maintain its 150MB/s throughput during the certify. FireWire is limited to 55MB/s.

ReplyQuote
Posted : 19/12/2017 12:26 am
(@hbomb28)
Active Member Customer

FireWire is way too slow. You need a faster bus!

Over FireWire, 3-4 days each seems fast, even. Thunderbolt will allow a drive to maintain its 150MB/s throughput during the certify. FireWire is limited to 55MB/s.

ReplyQuote
Posted : 21/03/2018 5:21 pm
(@hbomb28)
Active Member Customer

Hi, on a somewhat related note - i've run a certify single pass on a suspected bad disk, and it did indeed fail about 3/4 of the way through. i'd like to send it back to the manufacturer as it's still under warranty, but also want to make sure the data that was initially on it (before certifying) on it is NOT recoverable or viewable. I should also mention that this disk was initially one of 4 disks in a RAID5 volume.

Is it safe to assume that the softraid certify process effectively performs what mac os's 'zero out' disk utility feature does (ie writes zeros to each block so that previous data is overwritten)? or should i attempt to perform this 'zero out' specifically before sending it out? also, should i be worried about the fact that the certification process couldn't finish writing to all sectors? or is this all moot since the disk was part of a raid 5 volume and the chances of the info on it being rebuilt would be really difficult anyway?

thanks!

ReplyQuote
Posted : 21/03/2018 5:28 pm
(@softraid-support)
Member Admin

A certify wipes the disk. The last pass is zero's, each prior pass is a random number. That is why a 3 pass is most effective.

When your disk certified and failed 75% of the way through, if it failed on the write, then the last 25% of the drive has some data. However, that data is 1/4 of 64K, in 16K chunks. Some text might be recoverable, but no "binary" files.

If the drive failed on the "read" pass, then zero's were written across the entire disk. A "forensic recovery" by the FBI, for example can often get data off such a disk that was only "zero'd", but again, it will be 16K chunks, our of 64K sectors of data. Not much to go on.

That is why we recommend a 3-pass certify. With a modern drive, we believe it is most likely impossible to recover ANY data from a disk which has had a 3-pass certify on it.

ReplyQuote
Posted : 21/03/2018 9:39 pm
(@calbear88)
Eminent Member Customer

Hi, I was wondering, if I don't certify a disk, and the disk has a bad sector, would the bad sector be detected later on when I use up space on the disk and an attempt made to write to a bad sector fails?

I think that certifying marks which sectors on a disk is bad, but I was wondering if softraid or the operating system would do the same thing later on when writing to that sector, assuming I eventually use up all the space on the hard drive.

Thanks

ReplyQuote
Posted : 16/10/2018 9:57 pm
(@softraid-support)
Member Admin

Here is the problem. Disks are very different than in the 80's and 90's. A single bad sector now is a predictor of disk failure in a few weeks, or best case months.

So you do not want ANY bad sectors (reallocated sectors, essentially)

When you get a new disk, there may be bad areas in the disk. if you do not discover them until a year has passed, you end up with a disk that should have been "DOA" and returned, rather than a bum disk you did not discover the problem for a year.

Also, after 30 days, a drive manufacturer will replace a disk with a "refurbished" disk. You do not really want this either.

So best is get new disks, certify them to eliminate those disks which are defective on arrival. this means you have eliminated about 30% to 50% of your first years failures (or longer) in a matter of a week.

Disks can automatically reallocate sectors. However, another factor is the process can take up to two minutes, whereupon the computer will hang/crash. You do not want that either.

ReplyQuote
Posted : 17/10/2018 2:16 am
Share:
close
open