SSDs - State of the Product?

sechs

Storage? I am Storage!
Joined
Feb 1, 2003
Messages
4,709
Location
Left Coast
They're SSDs. They're legitimately faster. And they're what's available now.
Previous generation SSDs appear to be available and cheaper. Granted, they're probably using the smaller flash; but few people are going to get to the end of their drive's life before it gets replaced.

I guess that begs the question, when will people start putting together server farms of questionable SSDs running ZFS or the like?
 

ddrueding

Fixture
Joined
Feb 4, 2002
Messages
19,728
Location
Horsens, Denmark
I'm just putting regular SSDs in servers and running them with good backups. Haven't had any of those go bad yet. Some are getting hit quite hard; VMs with SQL databases on them and such.

You need a good backup strategy anyway, what is the big deal?
 

Handruin

Administrator
Joined
Jan 13, 2002
Messages
13,926
Location
USA
I could see a potential big deal for several reasons. One, you're constantly restoring backups and productivity is less because of constantly unreliable drives (theoretical concern). If you run with these in an array, your array will be in a degraded state and/or rebuilding more frequently causing a reduction in performance. Aside from that, I understand your point. Use the drive with the expectation that it's expendable and just enjoy the benefit for now.
 

Bozo

Storage? I am Storage!
Joined
Feb 12, 2002
Messages
4,396
Location
Twilight Zone
With the speed of SSDs, would you even notice a drop in performance if one died in an array? ...assuming a good controller...
 

Santilli

Hairy Aussie
Joined
Jan 27, 2002
Messages
5,278
Odd that you pick an absolute fanatic for SSD's, who wouldn't care if they failed every 12 months, he'd still use them:confused:?

Have to check out the new model he recommends...
 

ddrueding

Fixture
Joined
Feb 4, 2002
Messages
19,728
Location
Horsens, Denmark
In response to your blind faith in Ted Talks, I give you: Solid state drives fail. A lot. ;)

The total count he quotes is 11 drives. My personal collection is larger than that; not including the ones I've sold and used for work. If you are going to bring anecdotal evidence to challenge my anecdotal evidence, at least make it a larger sample.

And what's up with the "blind faith" jab? They're interesting.
 

sechs

Storage? I am Storage!
Joined
Feb 1, 2003
Messages
4,709
Location
Left Coast
Alright, I found one reason to get new-gen drives: FLAC encoding.

My previous-gen Sandforce-based drive still can't max out all six cores on my computer while encoding multiple FLACs. It's still drive-bound.

Of course, like booting, FLAC encoding isn't something that I do all day, every day....
 

time

Storage? I am Storage!
Joined
Jan 18, 2002
Messages
4,932
Location
Brisbane, Oz
With the speed of SSDs, would you even notice a drop in performance if one died in an array? ...assuming a good controller...

Under mulithreaded load, yes, because the drives can no longer service reads concurrently (assuming RAID 5). But for sequential transfers, probably no obvious difference.

The big advantage is that rebuilds should be many times faster than with spinning disks. Intel talks about SSD RAID here.
 

time

Storage? I am Storage!
Joined
Jan 18, 2002
Messages
4,932
Location
Brisbane, Oz
And what's up with the "blind faith" jab? They're interesting.

Sorry, didn't mean to offend. I'm currently without speakers while I try to repair my desk, so it's too much hassle to listen to that talk; I'll try to catch it later.

However, I disagree with the thesis implied by the description. Although you could construct psychological experiments to support it, I think you'll find that in practice, the opposite is more often true.

Caution stems from knowledge of possible consequences, not ignorance. It's summed up in the oft-quoted line by Alexander Pope, "Fools rush in where angels fear to tread".
 

ddrueding

Fixture
Joined
Feb 4, 2002
Messages
19,728
Location
Horsens, Denmark
No worries, Time. If I was without speakers, I'd be cranky too ;)

I have to disagree with you on the "familiar vs. unfamiliar risk", though.

Driving a car: people are familiar and therefore underestimate the risk.
Flying a plane: people are less familiar and therefore overestimate the risk.
Smoking: People do it every day, and many don't appreciate the consequences.

An earthquake is a pretty good example. Anyone who has been through a couple knows they are no big deal, but many who haven't are convinced those of us who live near faults are crazy.

Or my wife, who is terrified of me flying small aircraft. I hope to remedy that tomorrow by taking her up for her first single-piston flight.
 

Handruin

Administrator
Joined
Jan 13, 2002
Messages
13,926
Location
USA
This is one of my favorites...not really related

y7Hm9.jpg
 

Stereodude

Not really a
Joined
Jan 22, 2002
Messages
10,865
Location
Michigan
Alright, I found one reason to get new-gen drives: FLAC encoding.

My previous-gen Sandforce-based drive still can't max out all six cores on my computer while encoding multiple FLACs. It's still drive-bound.

Of course, like booting, FLAC encoding isn't something that I do all day, every day....
Were you reading and writing to the same drive? What program are you using? What format are the source files?
 

LunarMist

I can't believe I'm a Fixture
Joined
Feb 1, 2003
Messages
17,497
Location
USA
An earthquake is a pretty good example. Anyone who has been through a couple knows they are no big deal, but many who haven't are convinced those of us who live near faults are crazy.

You probably have not had your home completely destroyed or suffered bodily injury from falling objects. In 1971 I was hit in the head with something that weighed about 3 lbs. Not fun.

Why is this all in the SSD thread?
 

sechs

Storage? I am Storage!
Joined
Feb 1, 2003
Messages
4,709
Location
Left Coast
Were you reading and writing to the same drive? What program are you using? What format are the source files?
I tried it both ways. Obviously it's faster drive to drive than drive to self; but, even then, not more than four cores was maxed out at a time. The max speed was the same either way.

I was using foobar2000 to convert FLAC-to-FLAC.
 

ddrueding

Fixture
Joined
Feb 4, 2002
Messages
19,728
Location
Horsens, Denmark
You probably have not had your home completely destroyed or suffered bodily injury from falling objects. In 1971 I was hit in the head with something that weighed about 3 lbs. Not fun.

I live (and did at the time) ~30 miles from the epicenter of the Loma Prieta Earthquake. I was hit with some glass from a blown-out window, but nothing weighing 3lbs. Still 22 years ago we had something that killed less people than a number of tornadoes this month, or even the normal snowy weather on the east coast.

Why is this all in the SSD thread?

People being more afraid of things they don't understand and are unfamiliar with. Then it went a bit long.
 

Stereodude

Not really a
Joined
Jan 22, 2002
Messages
10,865
Location
Michigan
I tried it both ways. Obviously it's faster drive to drive than drive to self; but, even then, not more than four cores was maxed out at a time. The max speed was the same either way.

I was using foobar2000 to convert FLAC-to-FLAC.
FWIW, I don't have problem doing FLAC to FLAC with a 7200RPM HD (same drive) with 4 cores (Q6600 @ 3.0gHz). :cool:
 

time

Storage? I am Storage!
Joined
Jan 18, 2002
Messages
4,932
Location
Brisbane, Oz
That had nothing to do with a nuclear reactor of ancient design that was badly built.

That's news to me. Where did someone say it was badly built? And what aspect of the design that contributed to the failures has changed in more recent water-cooled designs?
 

ddrueding

Fixture
Joined
Feb 4, 2002
Messages
19,728
Location
Horsens, Denmark
I'll try to find it again, but the story I read was that an error while pouring the concrete was covered up.

Edit: The newest reactor designs can go completely cold while dry. I would consider that a mandatory safety feature.
 

Mercutio

Fatwah on Western Digital
Joined
Jan 17, 2002
Messages
22,275
Location
I am omnipresent
If the drive is really dead, how would they know?

Every Intel thing I've sent back ever, they send me an Email that indicates that the defective product has been examined and found to be covered by its warranty or not. In one case I sent in a motherboard where I was told that the CPU socket had two bent pins, so they never checked the issue (dead onboard NIC) that caused me to RMA the board in the first place.
 

LunarMist

I can't believe I'm a Fixture
Joined
Feb 1, 2003
Messages
17,497
Location
USA
Every Intel thing I've sent back ever, they send me an Email that indicates that the defective product has been examined and found to be covered by its warranty or not. In one case I sent in a motherboard where I was told that the CPU socket had two bent pins, so they never checked the issue (dead onboard NIC) that caused me to RMA the board in the first place.

I meant that if the drive were really dead the E9 attribute would not be readable.
 
Top