• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Does anyone know if the 7K3000 is emulating 512B sectors?

G

Gilbo

n00b
Joined
Feb 17, 2011
Messages
5
It's described as a 512B sector drive on the Hitachi product page, and I know odditory believes they're native 512 B sectors (and he's got a point about the 5-platter, lower-density design maybe requiring less ECC), but I've seen contradicting information.


Anyone know for sure?

Anyone done 4K random write tests with an unaligned and an aligned partition to check if performance is different?


Lastly, 1st post here. Thanks for having me!

I'm concerned because I may use these with ZFS, and I know that the way ZFS writes data to drives means if a drive is advertising an untrue sector size, you're going to end up with issues.
 
Thanks Cyberbeing. It does look like they would mention in that document if 512 bytes was an emulated sector size, and they don't.

It's nice to know there are a few, large-capacity drives out there that aren't emulating 512 byte sectors.


I wouldn't mind a 4K sector size, as long as the drive advertised the correct size, but I don't think there any of those available yet due to the fact EFI motherboards are only just becoming more common place.
 
StorageReview probably just got confused by Hitachi 3TB FAQ PDF:
Hitachi 3TB FAQ said:
How do external storage providers work around the 2.2TB limit?
To solve the 2.2TB limit, some external USB drives will present themselves with a sector size of 4K. By transferring 4K blocks, the USB drive effectively reduces the number of blocks that appear to the OS. Legacy USB drivers support 4K block sizes and will continue to work correctly. Although this implementation will work well for USB devices (which use USB drivers), it will not work for SATA devices, such as bare hard drives, since the SATA drivers only support 512‐byte transfers.
Click to expand...
This just means that with 3TB external drives, USB drivers are capable of 4096-byte streaming reads/writes (eight 512-byte sectors at once) to overcome the 2.2TB limitation on 32bit systems.

As far as I know, Hitachi only uses Advanced Format on some of their TravelStar (laptop) HDDs.
 
Last edited:
Don't know about any blah blah, but my 3TB 7K3000 are flying on RAID50 :D
 
Not having 4K sectors might actually be a disadvantage since you may need more redundancy to take over the function of the ECC on the HDD. The ECC on the HDD should make sure it can read all sectors but if it fails in this task, then you would have to 'borrow' redundancy just to make the disk remember all that it stored. That means that RAID-Z will no longer protect you from a single failure and RAID-Z2 won't protect you from a double failure. But rather RAID-Z2 will protect for one full failure and another disk that encounters uBER.

If you can live with some files dying on the volume, then this wouldn't be that big of a problem. ZFS metadata is always replicated (ditto blocks) and as such only files could become corrupt due to BER. A common pitfall is both a complete failure of one disk (in a RAID-Z for example) and while rebuilding/syncing one of the remaining disks encounters uBER. Then you are in trouble, and in those circumstances having 4KiB sector disks would certainly help as the disk itself could correct more bit errors and wouldn't have to rely on redundancy.

I would focus on high sequential speeds and lowest rpm you can get. The IOps you can acheive with SSD as SLOG+L2ARC (Intel G3/ Marvell C400 recommended). The high data density + 4K sectors would give 5400rpm and other low spindle HDDs alot of sequential performance.

So for the future 4200rpm doing 200MB/s and SSDs for IOps would be the dream; then you could get the best out of both technologies. For now looking for 5400rpm/7200rpm would seem a more conventional choice. But often older 512-byte sector 7200rpm HDDs actually perform worse than 4K sector 5400rpm disks, so don't be too afraid about 4K sectors!
 
sub.mesa said:
Not having 4K sectors might actually be a disadvantage since you may need more redundancy to take over the function of the ECC on the HDD. The ECC on the HDD should make sure it can read all sectors but if it fails in this task, then you would have to 'borrow' redundancy just to make the disk remember all that it stored. That means that RAID-Z will no longer protect you from a single failure and RAID-Z2 won't protect you from a double failure. But rather RAID-Z2 will protect for one full failure and another disk that encounters uBER.
Click to expand...
Whiskey Tango Foxtrot? :confused:

Aside from alignment issues there is no difference. Neither is more reliable than the other, or better protected, or offers more redundancy.
 
Not redundancy, but it would offer better protection against uBER; that is the whole point of 4K sectors. Current high capacity 2TB+ drives require more error correction to prevent issues with uBER. 4K sector drives can help with that, since it offers more error correction per sector.

So my point was that, when using 4K disks you would be better protected against uBER in the case of a RAID5 / RAID-Z with one failed disk. During the rebuild you don't want to encounter uBER or that will translate to inaccessible / corrupt files.

So when using high capacity drives with 512 byte sectors i would opt for more redundancy, like RAID-Z2. You could consider that RAID-Z2 to offer the protection of one complete disk failure + one disks encountering uBER. So just to protect against uBER would would want to reserve one level of redundancy.
 
sub.mesa said:
Not redundancy, but it would offer better protection against uBER; that is the whole point of 4K sectors. Current high capacity 2TB+ drives require more error correction to prevent issues with uBER. 4K sector drives can help with that, since it offers more error correction per sector.
Click to expand...
Uh... strictly in terms of ECC bits per sector 4k sector drives offer less protection. However, since the algorithms are more advanced and the sectors are larger you get equivalent protection with less bits used for the internal ECC. This means less of the raw space on the drive is "wasted" on error correction.

Again, it has no impact on UBE's or the like. It simply allows the HD makers to get better usable space efficiency against the raw bit space on the platters.
 
Stereodude said:
Uh... strictly in terms of ECC bits per sector 4k sector drives offer less protection. However, since the algorithms are more advanced and the sectors are larger you get equivalent protection with less bits used for the internal ECC. This means less of the raw space on the drive is "wasted" on error correction.
Click to expand...

Theoretically, you could save on percentage of drive used for ECC when switching from 512B to 4KB sectors, if you rewrote your firmware for a sufficiently sophisticated ECC method. Alternatively, you could theoretically achieve better UBER while using the same space for ECC. Or somewhere in between these extremes.

However, I have seen no evidence that the manufacturers have yet done this. It could be that so far, the HDD manufacturers have simply implemented a simple firmware patch to split/combine 512B sectors into 4KB sectors, while utilizing the same ECC algorithms as for 512B sectors. The reason I think this is likely is because all of the HDD manufacturers are reporting the same usable capacities and UBER rates for their equivalent models of physical-512B- and physical-4KB-sector drives.

If the manufacturers have not yet implemented more advanced ECC with physical-4KB-sector drives, it is likely that they will be forced to do so in the future as platter densities, and the associated raw error rates, continue to increase.

But at present, if you examine the spec sheets for physical-4KB-sector HDDs, the UBER is no better than for equivalent 512B sector models. In fact, I recently noticed that Western Digital had quietly lowered the UBER spec for their entire line of green drives, including physical-4KB-sector models, from 1 in 10^15 to 1 in 10^14. I'm not sure why, but clearly the 4KB-physical-sector drives have not improved UBER for them.
 
sub.mesa said:
Not redundancy, but it would offer better protection against uBER; that is the whole point of 4K sectors. Current high capacity 2TB+ drives require more error correction to prevent issues with uBER. 4K sector drives can help with that, since it offers more error correction per sector.

So my point was that, when using 4K disks you would be better protected against uBER in the case of a RAID5 / RAID-Z with one failed disk. During the rebuild you don't want to encounter uBER or that will translate to inaccessible / corrupt files.

So when using high capacity drives with 512 byte sectors i would opt for more redundancy, like RAID-Z2. You could consider that RAID-Z2 to offer the protection of one complete disk failure + one disks encountering uBER. So just to protect against uBER would would want to reserve one level of redundancy.
Click to expand...

Hate to disagree since you're right 99% of the time, and I don't want to use the word FUD but this comes close. The reason the Hitachi 7K3000 3TB gets away with 512 byte native sectors is because of its 5 platter design, plain and simple. Which means obviously lower density per surface and higher SNR which Hitachi obviously felt didn't require the more powerful ECC afforded by 4K sectors. Competing 2TB and 3TB drives felt they had to go 4K because of their higher surface density.

So sure 4K sectors offer more powerful ECC but as we know there are numerous headaches to 4K drives for many people, not everyone wants to run FreeBSD+zfs or hardware raid. Many people are still on systems that have problems with 4K sectors and will be for years to come - I mean non tech-savvy end users.

In all likelihood the 5 platter 3TB drive will be the last 512 byte form factor from Hitachi before they're forced to move to 4K.
 
Last edited:
Stereodude said:
Uh... strictly in terms of ECC bits per sector 4k sector drives offer less protection.
Click to expand...
But... does it really work like that? I know the 512-byte sectors have 40-byte ECC and 4K sectors have 80-byte ECC. So you're right that there's less bits of ECC per bits of data. But does that mean less protection?

I was under the impression that ECC works like:
Can correct up to X bit errors
Can detect up to Y bit errors, where Y > X.

With more ECC bits, that would translate to more bit errors; though we have more data and thus possibly more errors to correct. But still, this could recover data that is unrecoverable with less ECC.

The reports i've got from people also seem to agree, though that in itself is no proof as other factors can be involved. The issues with WD Green 2TB with 512-byte sectors are alot of Current Pending Sectors where if you correct them by writing data to it, it would not raise the Reallocated Sector Count; so the sectors themselves were fine; no physical damage. Just the ECC could not recover the data; more ECC would have helped here. Quite a few people i've helped on other boards with WD Green and current pending sectors. Some has thousands of these, with NO reallocated sector counts even after the zero write.

I've not yet received such reports from Samsung F4. Yes a few people with current pending sectors, but not many and when zero written they raised the reallocated sector count so they were true bad sectors. Didn't have many reports about EARS though, strange since that's quite a popular drive as well. These people were not using ZFS though; just casual windows users.

I agree this is hardly scientific though. If untrue, i would stand corrected. But is there any document you know of that has done an independent (i.e. not some HDD vendor) research about uBER specially in relation to 512B/4K sectors? Or research that compares measured uBER rate of 512-byte versus 4K disks? That would be very helpful. Unfortunately, i've found reliable and independent information to be scarce. And i do not trust the specifications that HDD makers put up on their site either. It's also a crude measure, in factors of 10. Actual measurements by independent testing would be very valuable information on this topic.
It simply allows the HD makers to get better usable space efficiency against the raw bit space on the platters.
Click to expand...
You're totally right that this is the prime reason for going 4K sectors, unlike what i earlier stated. As i understand, it's not just the ECC space benefit (less ECC bits 'wasted' per gigabyte storage), but primarily the SYNC/DAM which is like a 'lead-in' and 'lead-out' to a sector. With fewer sectors the relative space these blocks require/waste is cut by 7/8 (8 since 4K/512B = 8).
 
You must log in or register to reply here.
Back
Top