On 3/11/2022 6:31 am, Paul wrote:
>
> Stay away from the top capacity drives, unless you are sure
> of the specs. Some of them are "Host Managed" and the
> controller does not work the same way as a consumer drive.
> So at 20TB max today, you would be relatively safe buying
> 14TB NAS drives. Those should have ordinary controllers. For
> the most part, the consumer ones will be 512e (and not the
> useless 4k ones). The 4k ones cause tool or workflow issues.
> Newegg used to sell some, but they may have stopped based
> on the rate of return.

Doesn't make sense to manufacture host-based hard disks, does it? Some
simple Window$ bug would crash the drive leading to data loss...

On 11/3/2022 1:47 PM, Mr. Man-wai Chang wrote:
> On 3/11/2022 6:31 am, Paul wrote:
>>
>> Stay away from the top capacity drives, unless you are sure
>> of the specs. Some of them are "Host Managed" and the
>> controller does not work the same way as a consumer drive.
>> So at 20TB max today, you would be relatively safe buying
>> 14TB NAS drives. Those should have ordinary controllers. For
>> the most part, the consumer ones will be 512e (and not the
>> useless 4k ones). The 4k ones cause tool or workflow issues.
>> Newegg used to sell some, but they may have stopped based
>> on the rate of return.
>
> Doesn't make sense to manufacture host-based hard disks, does it? Some simple Window$ bug would crash the drive leading to data loss...

Dropbox (cloud provider) uses it.

https://blog.westerndigital.com/host-managed-smr-dropbox/

We would need to see the details of implementation, to see
if it made sense or not.

I haven't even seen a block diagram of what is on the
controller board for the disk drive. It may be more highly
integrated than we think.

It's all very secret stuff... because Dropbox pays for
development, not Seagate or WD. Dropbox writes the code.

BackBlaze, I think it handles faults at the drive level
or at the pod level. To take a pod out of service, they might
have to replicate it (if it still worked well enough for that
to function). At some point, there will be a fault group which
is larger than a single drive. Any rack mount with 25 drives
in it, could have a flaw in the motherboard that needs repair.

Paul

On 4/11/2022 2:29 am, Paul wrote:
> On 11/3/2022 1:47 PM, Mr. Man-wai Chang wrote:
>>
>> Doesn't make sense to manufacture host-based hard disks, does it? Some simple Window$ bug would crash the drive leading to data loss...
>
> Dropbox (cloud provider) uses it.
>
> https://blog.westerndigital.com/host-managed-smr-dropbox/
>
> We would need to see the details of implementation, to see
> if it made sense or not.
>
> I haven't even seen a block diagram of what is on the
> controller board for the disk drive. It may be more highly
> integrated than we think.
>
> It's all very secret stuff... because Dropbox pays for
> development, not Seagate or WD. Dropbox writes the code.

Thanks!

On 4/11/2022 2:29 am, Paul wrote:
> On 11/3/2022 1:47 PM, Mr. Man-wai Chang wrote:
>> On 3/11/2022 6:31 am, Paul wrote:
>>
>> Doesn't make sense to manufacture host-based hard disks, does it? Some simple Window$ bug would crash the drive leading to data loss...
>
> Dropbox (cloud provider) uses it.
>
> https://blog.westerndigital.com/host-managed-smr-dropbox/
>
> We would need to see the details of implementation, to see
> if it made sense or not.
>
> I haven't even seen a block diagram of what is on the
> controller board for the disk drive. It may be more highly
> integrated than we think.
>
> It's all very secret stuff... because Dropbox pays for
> development, not Seagate or WD. Dropbox writes the code.

BTW, reminded me of some old terms in computer science: fault-tolerant
real-time distributed system. :)

In alt.comp.hardware.pc-homebuilt Mr. Man-wai Chang <toylet.toylet@gmail.com> wrote in part:
> On 4/11/2022 2:29 am, Paul wrote:
>> On 11/3/2022 1:47 PM, Mr. Man-wai Chang wrote:
>>> On 3/11/2022 6:31 am, Paul wrote:
>>> Doesn't make sense to manufacture host-based hard disks, does it?
>>> Some simple Window$ bug would crash the drive leading to data loss...
>>
>> Dropbox (cloud provider) uses it.
>>
>> https://blog.westerndigital.com/host-managed-smr-dropbox/
>>
>> We would need to see the details of implementation, to see
>> if it made sense or not.
>>
>> I haven't even seen a block diagram of what is on the
>> controller board for the disk drive. It may be more highly
>> integrated than we think.
>>
>> It's all very secret stuff... because Dropbox pays for
>> development, not Seagate or WD. Dropbox writes the code.
>
> BTW, reminded me of some old terms in computer science:
> fault-tolerant real-time distributed system. :)

Or RAID -- Redundant Array of Inexpensive Disks. Host or HW based.
Perhaps the wheel is being reinvented yet again.

-- Robert

On 11/7/2022 1:34 PM, Robert wrote:
> In alt.comp.hardware.pc-homebuilt Mr. Man-wai Chang <toylet.toylet@gmail.com> wrote in part:
>> On 4/11/2022 2:29 am, Paul wrote:
>>> On 11/3/2022 1:47 PM, Mr. Man-wai Chang wrote:
>>>> On 3/11/2022 6:31 am, Paul wrote:
>>>> Doesn't make sense to manufacture host-based hard disks, does it?
>>>> Some simple Window$ bug would crash the drive leading to data loss...
>>>
>>> Dropbox (cloud provider) uses it.
>>>
>>> https://blog.westerndigital.com/host-managed-smr-dropbox/
>>>
>>> We would need to see the details of implementation, to see
>>> if it made sense or not.
>>>
>>> I haven't even seen a block diagram of what is on the
>>> controller board for the disk drive. It may be more highly
>>> integrated than we think.
>>>
>>> It's all very secret stuff... because Dropbox pays for
>>> development, not Seagate or WD. Dropbox writes the code.
>>
>> BTW, reminded me of some old terms in computer science:
>> fault-tolerant real-time distributed system. :)
>
> Or RAID -- Redundant Array of Inexpensive Disks. Host or HW based.
> Perhaps the wheel is being reinvented yet again.
>
> -- Robert
>

No, the wheel isn't being reinvented.

the drive type is SMR -- Shingled Magnetic Recording.
Done in seven track chunks.

The first generation of SMR drives for consumers,
they might read and write at 120MB/sec, but every once
in a while, the drive would "stutter" and it would write
at 25MB/sec (slow!). People of course, hated this.

A second generation of SMR drive, used the cache DRAM on
the controller, and additional cache behaviors were added,
to smooth out the writing. (To update one byte in a chunk,
requires reading seven tracks, changing the byte, writing
back the seven tracks, in one "smooth operation". You cannot
just write one single sector, on an SMR.)

The second generation managed to achieve read at 120MB/sec
and write at 110MB/sec (the average write rate on an SMR
drive is always less than the read rate, unlike
PMR [Perpendicular] drives).

This solution, of Dropbox writing code to do internal disk
drive operations, is an attempt to get the very best
behavior from "inferior goods" -- the SMR drive. The host
computer memory is used for caching. Multiple contiguous
7 track chunks may be written in one sequence.

*******

We can detect SMR drives, via the platter count. The platter
count is no longer in the datasheet. Neither is the word "SMR"
in a datasheet. All the details needed to make an informed
purchase, are not in the datasheet.

https://en.wikipedia.org/wiki/Shingled_magnetic_recording

"Western Digital, Toshiba, and Seagate have sold SMR drives without
labeling them as such, generating a large controversy, as SMR drives
behave much more slowly under some circumstances (such as random writes)
than PMR drives.[14] Some have even claimed that these may cause
data loss.[15] These mislabeling practices were used in both
consumer-centric, and dedicated data storage HDDs for servers,
NASes, RAIDs, and cold storage. A United States class-action suit
against Western Digital alleging the technology is inferior was
settled on or before August 27, 2021.[16]"

Just be careful when you are shopping. Don't buy crap.

Paul

On 8/11/2022 3:27 am, Paul wrote:
> https://en.wikipedia.org/wiki/Shingled_magnetic_recording
>
> "Western Digital, Toshiba, and Seagate have sold SMR drives without
> labeling them as such, generating a large controversy, as SMR drives
> behave much more slowly under some circumstances (such as random writes)
> than PMR drives.[14] Some have even claimed that these may cause
> data loss.[15] These mislabeling practices were used in both
> consumer-centric, and dedicated data storage HDDs for servers,
> NASes, RAIDs, and cold storage. A United States class-action suit
> against Western Digital alleging the technology is inferior was
> settled on or before August 27, 2021.[16]"
>
> Just be careful when you are shopping. Don't buy crap.

I don't remmeber any newspapers talking about this court case....