10TB Hard Drive, can't even be accessed by modern OS's yet!

"Charlie" wrote in message
...
On 3/14/2015 7:11 PM, Gene E. Bloch wrote:
On Sat, 14 Mar 2015 05:39:04 -0700, Drew wrote:

This was going to be my reply:

MY PLAN
Why would a home user need 10,000 gigabytes of storage? By the time you
fill it something new would come along or it

or the user

would die first.
/MY PLAN

but before doing it I read the other replies.

Obviously I didn't realize what some users' needs are.

Just think of what a backup would take!

I don’t bother to back it all up, essentially because its
spread over multiple physical drives many of which aren't
even plugged in so you'd only lose part of the total.

wrote in message
...
On Sun, 15 Mar 2015 04:38:39 -0400, Charlie wrote:

On 3/14/2015 7:11 PM, Gene E. Bloch wrote:
On Sat, 14 Mar 2015 05:39:04 -0700, Drew wrote:

This was going to be my reply:

MY PLAN
Why would a home user need 10,000 gigabytes of storage? By the time you
fill it something new would come along or it

or the user

would die first.
/MY PLAN

but before doing it I read the other replies.

Obviously I didn't realize what some users' needs are.

Just think of what a backup would take!

Another 10T disk? Unless I messed up the arithmetic, that's about a
day or two to do a complete backup.

But you wouldn't do backup like that, just write new
PVR files to both drives when a new one shows up.

On Sun, 15 Mar 2015 11:58:32 -0400, Paul wrote:

Char Jackson wrote:
On Sun, 15 Mar 2015 04:38:39 -0400, Charlie wrote:

On 3/14/2015 7:11 PM, Gene E. Bloch wrote:
On Sat, 14 Mar 2015 05:39:04 -0700, Drew wrote:

This was going to be my reply:

MY PLAN
Why would a home user need 10,000 gigabytes of storage? By the time you
fill it something new would come along or it
or the user

would die first.
/MY PLAN

but before doing it I read the other replies.

Obviously I didn't realize what some users' needs are.

Just think of what a backup would take!

I have a total of about 50TB here, consisting of two volumes, (using Drive
Bender), and made the decision long ago that I wouldn't be backing up all of
it under any circumstances. Instead, you decide what needs to be backed up
and do it selectively, and the rest either gets protected with parity so
that you can lose a drive or two and still recover, or you simply decide to
ride bareback and treat the data as expendable or replaceable.

How unhappy would you be, if you lost the entire array ?

Common mode failures do happen.

All it takes is a power supply failure, the 12V rail rising
to 15V for around 30 seconds, and it's all over for your array.

Well, your house could burn down, too (and much more likely). Just
unplug an array element and take it to another site. Rotate that
drive though the array with another, or three.

*******

One problem I see with that 10TB drive, is it's not going
to fit into the typical IT guys "backup window value".
You'd be surprised how important that is to some people.

I'm also surprised there's no "reach" program at Seagate
or WD, to change the basics of hard drive design. And
crank up the bandwidth. If you're going to make a 10TB
drive, it should have 500MB/sec bandwidth. They should
at least have enough heads, to write the entire shingle
in one pass.

On 03/15/2015 08:25 AM, Yousuf Khan wrote:

Maybe the special write procedures for these drives can't be fully
handled by the drive firmware, and so it needs an assist by the OS? It
wouldn't surprise me if the write timing operation is so complex that
it's best handled by a subroutine that can only run on the host machine.
The host machine would have better understanding of the high-level file
system data structures, which the firmware wouldn't understand. Perhaps
it's best to not consider these hard drives but something between a hard
drive and a tape drive?

http://storageconference.us/2014/Pre...ions/Novak.pdf

suggests that ZFS is the best file system for SMR drives.


Here's a piece about the Seagate SMR drive (only 8TB so far, I think):

http://open-zfs.org/w/images/2/2a/Ho...im_Feldman.pdf

Perce

On 2015-03-14 00:36, Yousuf Khan wrote:

There are no longer any gaps in-between each track which allows more
data to be stored on a single platter, but at the cost of more
complicated software on the OS to properly read, write, and over-write
data without destroying neighboring tracks.

That sounds strange; firmware should handle writes transparently, I
wonder what they mean...

--
! _\|/_ Sylvain /
! (o o) Member-+-David-Suzuki-Foundation/EFF/Planetary-Society-+-
oO-( )-Oo Just got a new computer for my wife. What a great trade!

On 2015-03-14 12:24, Paul wrote:

Paul wrote:
Wolf K wrote:

Why can't a 10TB HDD be configured the same way?

One of the commenters at the end of that Gizmodo
article, points out the difference.

http://gizmodo.com/sadly-this-10tb-h...not-1691245306


"There are three ways to implement SMR data management.

HGST uses host managed, which is perfect as they're targeting
hyperscale with these drives.

Seagate uses drive managed, which works with any OS/file system.
Seagate is going for more wide reaching market penetration where
HGST is focused on top 5-10 guys only.

Incidentally, that large hyperscale market doesn't have a problem
with modifying their stack to accommodate SMR.
"

So it's a conscious design decision, to "let the user"
write the handler. In the case of the HGST product.

Paul

Darn, I replied too fast, here's my answer... ;-)

--
! _\|/_ Sylvain /
! (o o) Member-+-David-Suzuki-Foundation/EFF/Planetary-Society-+-
oO-( )-Oo "Apple" (c) Copyright 1767, Sir Isaac Newton.

B00ze/Empire wrote:
On 2015-03-14 12:24, Paul wrote:

Paul wrote:
Wolf K wrote:

Why can't a 10TB HDD be configured the same way?

One of the commenters at the end of that Gizmodo
article, points out the difference.

http://gizmodo.com/sadly-this-10tb-h...not-1691245306



"There are three ways to implement SMR data management.

HGST uses host managed, which is perfect as they're targeting
hyperscale with these drives.

Seagate uses drive managed, which works with any OS/file system.
Seagate is going for more wide reaching market penetration where
HGST is focused on top 5-10 guys only.

Incidentally, that large hyperscale market doesn't have a problem
with modifying their stack to accommodate SMR.
"

So it's a conscious design decision, to "let the user"
write the handler. In the case of the HGST product.

Paul

Darn, I replied too fast, here's my answer... ;-)

It's kinda unbelievable. Obviously HGST want to sell
a limited number of those. Maybe they'll analyze the
Seagate drives and figure out the best policies :-)

Paul