This is relevant - "Why solid-state disks are winning the argument".

On 2014-11-19 22:25:03 +0000, nospam said:

In article , Oregonian Haruspex
wrote:

I see no benefit to those new-fangled "hybrid" drives (really just a HD
with a bigger, smarter cache) because the rust will still be spinning
all the time.

hybrid drives are actually not that great and only slightly better than
a normal hd. it's basically a big cache for recently used files, which
may not be the ones that matter.

Yeah I know. Caching is a gamble anyway but especially when the OS and
the drive aren't talking to each other about it.

they don't need to. the drive cache holds recently accessed blocks with
the assumption they might be needed again.

As far as I understand it, the drive reads ahead and stores blocks in
the same sector, assuming that the OS might need them. Sequential
read-ahead. The OS is what stores the recently accessed blocks.

apple's fusion drive is a much better solution

It's the same exact thing as any other hybrid SSD but with more
marketing power behind it.

it's not the same thing at all.

fusion moves files between ssd and hd based on usage patterns. apps and
documents that are used frequently will end up on ssd and what is
rarely used will be on hd, completely automatically.

This is marketing speak. Are there any rechnical documents that
describe excatly what's happening? If not I don't like to believe
these kinds of things.

It's possible that they took the ARC concept from ZFS and applied it to
OS X, but that's just more OS-level caching and not the "fusion drive"
that's doing it.

On 2014-11-19 15:30:17 +0000, nospam said:

In article , Eric Stevens
wrote:

http://www.theregister.co.uk/2014/11/07/storage_ssds/

... Unless your workload is very specifically single source, massive
capture, then you should be running SSDs. Even if you are not running
pure SSD, the case for tiered or hybrid storage makes itself.

SSDs are faster. They have way lower latency. They consume less power.
They take up less space.

and way more reliable, even if you're hammering it.

This remains to be seen, over the long term. Though it seems logical
because of the lack of moving parts.

But capacitors these days are really ****ty so who knows?

I've only ever had one hard drive fail in my life. I got my first HD
back in 1988. I only have 2 SSDs in the house so we'll see.

In article , "Mayayana"
writes:

"Similarly, SSDs are a terrible place to do a bunch of log file writes to;
eleventy squillion crappy little sub-K writes will burn out the SSDs in no
time."

So an SSD might make a good D drive, but probably not
such a good C drive. (Though I don't actually know how
much a "squillion" is. No surprise that the author is not one
of the British regulars at The Register. As much as the
British like to use their own slang overly much, at least
they don't talk like children.

This is definitely an issue if you have lots of writes. I don't know
how many "lots" is; it will depend on the type of disk, and this will
improve with time. However, modern SSDs can burn out in half an hour on
real-world systems with lots of writes. For write-once or write-few, of
course, they are OK.

In article , Phillip Helbig (undress to
reply) wrote:

This is definitely an issue if you have lots of writes. I don't know
how many "lots" is; it will depend on the type of disk, and this will
improve with time. However, modern SSDs can burn out in half an hour on
real-world systems with lots of writes. For write-once or write-few, of
course, they are OK.

complete bull****.

http://arstechnica.com/gadgets/2014/...s-actually-las
t-a-hell-of-a-long-time/
The results are impressive: the consumer-grade SSDs tested all made
it to at least 700TB of writes before failing. Three of the drives
have written 1PB (that¹s a thousand terabytes, by TechReport¹s
decimal reckoning, not 1024TB). That¹s a hell of a lot more writes
than the manufacturers¹ stated drive lifetimes, and that¹s good news
for SSD-buying consumers.

...it took almost a year of solid torture-test writes to get them to
failure.

In article , Oregonian Haruspex
wrote:

I see no benefit to those new-fangled "hybrid" drives (really just a HD
with a bigger, smarter cache) because the rust will still be spinning
all the time.

hybrid drives are actually not that great and only slightly better than
a normal hd. it's basically a big cache for recently used files, which
may not be the ones that matter.

Yeah I know. Caching is a gamble anyway but especially when the OS and
the drive aren't talking to each other about it.

they don't need to. the drive cache holds recently accessed blocks with
the assumption they might be needed again.

As far as I understand it, the drive reads ahead and stores blocks in
the same sector, assuming that the OS might need them. Sequential
read-ahead. The OS is what stores the recently accessed blocks.

drives generally cache the entire file when one block is accessed
because there's an extremely high likelihood you'll be accessing the
entire file.

however, that's very different than what fusion does.

apple's fusion drive is a much better solution

It's the same exact thing as any other hybrid SSD but with more
marketing power behind it.

it's not the same thing at all.

fusion moves files between ssd and hd based on usage patterns. apps and
documents that are used frequently will end up on ssd and what is
rarely used will be on hd, completely automatically.

This is marketing speak.

it is not. fusion is very different than a simple cache.

Are there any rechnical documents that
describe excatly what's happening?

not really much more than what i said above.

If not I don't like to believe
these kinds of things.

that's up to you, but it's definitely not marketing speak.

It's possible that they took the ARC concept from ZFS and applied it to
OS X, but that's just more OS-level caching and not the "fusion drive"
that's doing it.

it's not the arc concept and fusion is done in core storage, which is
part of the os.

In article , Oregonian Haruspex
wrote:

and way more reliable, even if you're hammering it.

This remains to be seen, over the long term. Though it seems logical
because of the lack of moving parts.

it's already been seen, which is partly why ssd warranties are longer
than hard drives, as much as 10 years, while hard drives are usually 1
year and sometimes 3 or 5 years.

But capacitors these days are really ****ty so who knows?

capacitors are fine now. they were bad around a decade ago.

I've only ever had one hard drive fail in my life. I got my first HD
back in 1988. I only have 2 SSDs in the house so we'll see.

buy a lottery ticket.

i'm seeing approximately one drive failure every year or so.

On 2014-11-20 05:31:06 +0000, nospam said:

In article , Oregonian Haruspex
wrote:

I see no benefit to those new-fangled "hybrid" drives (really just a HD
with a bigger, smarter cache) because the rust will still be spinning
all the time.

hybrid drives are actually not that great and only slightly better than
a normal hd. it's basically a big cache for recently used files, which
may not be the ones that matter.

Yeah I know. Caching is a gamble anyway but especially when the OS and
the drive aren't talking to each other about it.

they don't need to. the drive cache holds recently accessed blocks with
the assumption they might be needed again.

As far as I understand it, the drive reads ahead and stores blocks in
the same sector, assuming that the OS might need them. Sequential
read-ahead. The OS is what stores the recently accessed blocks.

drives generally cache the entire file when one block is accessed
because there's an extremely high likelihood you'll be accessing the
entire file.

No they don't, because drives do not know the physical structure of the
filesystem nor what blocks in which a certain file is located! This is
the job of the OS. The drive just fetches blocks, and hopes that it
can fetch the right ones.

To my knowledge there are no drives existing that know about the
filesystems they contain.

however, that's very different than what fusion does.

The only thing we know about Fusion is from marketing material. Let's
not pretend we know more than we do now ok?

apple's fusion drive is a much better solution

It's the same exact thing as any other hybrid SSD but with more
marketing power behind it.

it's not the same thing at all.

fusion moves files between ssd and hd based on usage patterns. apps and
documents that are used frequently will end up on ssd and what is
rarely used will be on hd, completely automatically.

This is marketing speak.

it is not. fusion is very different than a simple cache.

Where is the technical documentation for Fusion? How do you know how it works?

Are there any rechnical documents that
describe excatly what's happening?

not really much more than what i said above.

Just because you read some marketing material, this doesn't make it
true and certainly you don't have to believe it.

If not I don't like to believe
these kinds of things.

that's up to you, but it's definitely not marketing speak.

How can you know for sure without the technical details?

It's possible that they took the ARC concept from ZFS and applied it to
OS X, but that's just more OS-level caching and not the "fusion drive"
that's doing it.

it's not the arc concept and fusion is done in core storage, which is
part of the os.

The details are sketchy and thin and it seems almost exactly like ARC
in ZFS, but because there's really no technical documentation on this
it's impossible to say.

I like to actually know what's going on before I say I know what's going on.

On 2014-11-20 05:31:08 +0000, nospam said:

In article , Oregonian Haruspex
wrote:

and way more reliable, even if you're hammering it.

This remains to be seen, over the long term. Though it seems logical
because of the lack of moving parts.

it's already been seen, which is partly why ssd warranties are longer
than hard drives, as much as 10 years, while hard drives are usually 1
year and sometimes 3 or 5 years.

But capacitors these days are really ****ty so who knows?

capacitors are fine now. they were bad around a decade ago.

They are getting smaller and smaller for a given capacity, and because
they are often surface mounted these days this means that you can't fix
one when it goes bad.

Nobody knows how long a new capacitor formulation will last, and nobody
knows what sorts of batch to batch and part to part variations are
present until something goes wrong.

In article , Oregonian Haruspex
wrote:

I see no benefit to those new-fangled "hybrid" drives (really just a HD
with a bigger, smarter cache) because the rust will still be spinning
all the time.

hybrid drives are actually not that great and only slightly better than
a normal hd. it's basically a big cache for recently used files, which
may not be the ones that matter.

Yeah I know. Caching is a gamble anyway but especially when the OS and
the drive aren't talking to each other about it.

they don't need to. the drive cache holds recently accessed blocks with
the assumption they might be needed again.

As far as I understand it, the drive reads ahead and stores blocks in
the same sector, assuming that the OS might need them. Sequential
read-ahead. The OS is what stores the recently accessed blocks.

drives generally cache the entire file when one block is accessed
because there's an extremely high likelihood you'll be accessing the
entire file.

No they don't, because drives do not know the physical structure of the
filesystem nor what blocks in which a certain file is located! This is
the job of the OS. The drive just fetches blocks, and hopes that it
can fetch the right ones.

To my knowledge there are no drives existing that know about the
filesystems they contain.

the os tells the drive what to read and it's cached.

however, that's very different than what fusion does.

The only thing we know about Fusion is from marketing material. Let's
not pretend we know more than we do now ok?

nobody is pretending. quite a bit has been written about it, most of
which is not from apple.

apple's fusion drive is a much better solution

It's the same exact thing as any other hybrid SSD but with more
marketing power behind it.

it's not the same thing at all.

fusion moves files between ssd and hd based on usage patterns. apps and
documents that are used frequently will end up on ssd and what is
rarely used will be on hd, completely automatically.

This is marketing speak.

it is not. fusion is very different than a simple cache.

Where is the technical documentation for Fusion? How do you know how it
works?

because i've read quite a bit about it.

here's apple's tech note:
http://support.apple.com/en-us/HT202574
Presented as a single volume on your Mac, Fusion Drive automatically
and dynamically moves frequently used files to Flash storage for
quicker access, while infrequently used items move to the hard disk.
As a result you'll enjoy shorter startup times, and as the system
learns how you work you'll see faster application launches and
quicker file access. Fusion Drive manages all this automatically in
the background.

the exact algorithms by which it moves stuff is not public, but none of
that matters.

as the user uses the computer, commonly used files are moved to the ssd
and infrequently used files are moved to the hard drive, without the
user needing to do anything other than use the computer normally.

Are there any rechnical documents that
describe excatly what's happening?

not really much more than what i said above.

Just because you read some marketing material, this doesn't make it
true and certainly you don't have to believe it.

see above.

If not I don't like to believe
these kinds of things.

that's up to you, but it's definitely not marketing speak.

How can you know for sure without the technical details?

see above and i trust those who have tested it (as opposed to simply
used it).

It's possible that they took the ARC concept from ZFS and applied it to
OS X, but that's just more OS-level caching and not the "fusion drive"
that's doing it.

it's not the arc concept and fusion is done in core storage, which is
part of the os.

The details are sketchy and thin and it seems almost exactly like ARC
in ZFS, but because there's really no technical documentation on this
it's impossible to say.

it's nothing like arc.

I like to actually know what's going on before I say I know what's going on.

yet you keep saying what it isn't, without knowing what's going on.

In article , Oregonian Haruspex
wrote:

and way more reliable, even if you're hammering it.

This remains to be seen, over the long term. Though it seems logical
because of the lack of moving parts.

it's already been seen, which is partly why ssd warranties are longer
than hard drives, as much as 10 years, while hard drives are usually 1
year and sometimes 3 or 5 years.

But capacitors these days are really ****ty so who knows?

capacitors are fine now. they were bad around a decade ago.

They are getting smaller and smaller for a given capacity, and because
they are often surface mounted these days this means that you can't fix
one when it goes bad.

so what?

and the capacitors that failed were electrolytics, not smd.

Nobody knows how long a new capacitor formulation will last, and nobody
knows what sorts of batch to batch and part to part variations are
present until something goes wrong.

what does that have to do with ssd? hard drives have capacitors too, as
do computers.

there is a backup capacitor on recent ssds in case of power failure but
that's really about the extent of it. older ssds without a capacitor
have a small risk of corruption in the event of a power failure but if
the ssd is in a laptop, it's a non-issue since the laptop's battery is
the backup power.