General HDDs Discussion

[nameruse]

**** Post for FREE via your newsreader at post.usenet.com ****

Hi all,

I think there are a number of factors such as noise, power,
shock resistance, performance, and heat, that are important
to HDD users.

Forever increasing the RPM cannot really be the answer to
better performance, can it? For a 15k rpm HDD, its access time
is around 6 ms, still about 1000~10000 times slower than
memory-based storage. If we move to solid state instead of
spinning platters, could we then expect frictional heat and
shock to become much less of a concern? How about using other
methods to suspend the R/W head instead of on air? A frictionless
and/or silent spinning motor design might also be interesting.

Also, for HDDs, I think:
energy required = energy doing actually work + noise + heat.

If we can remove/reduce the noise and heat, the power required
should also decrease accordingly. In view of current products,
it seems performance and (noise + heat) seem to go hand in hand

Maybe HDDs are not the best storage device after all? If cost is
not a concern, solid-state drives seem to be superior in every
aspect. What alternatives are available? For example, several
years ago I read that IBM had been developing holographic storage.
Is it still in development or commercialised?

Just a few thoughts.



-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
*** Usenet.com - The #1 Usenet Newsgroup Service on The Planet! ***
http://www.usenet.com
Unlimited Download - 19 Seperate Servers - 90,000 groups - Uncensored
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=

 

[Eric Gisin]

The best way to improve performance in a server is distribute IO over many
drives with RAID.

The new way to get more drives in the same space is the new Seagate Savio 2.5"
10K drive. You get more IO/s with two of these over a single 15K drive.

 

[Al Dykes]

**** Post for FREE via your newsreader at post.usenet.com ****

Hi all,

I think there are a number of factors such as noise, power,
shock resistance, performance, and heat, that are important
to HDD users.

Forever increasing the RPM cannot really be the answer to
better performance, can it? For a 15k rpm HDD, its access time
is around 6 ms, still about 1000~10000 times slower than
memory-based storage. If we move to solid state instead of
spinning platters, could we then expect frictional heat and
shock to become much less of a concern? How about using other
methods to suspend the R/W head instead of on air? A frictionless
and/or silent spinning motor design might also be interesting.

Also, for HDDs, I think:
energy required = energy doing actually work + noise + heat.

If we can remove/reduce the noise and heat, the power required
should also decrease accordingly. In view of current products,
it seems performance and (noise + heat) seem to go hand in hand

Maybe HDDs are not the best storage device after all? If cost is
not a concern, solid-state drives seem to be superior in every
aspect. What alternatives are available? For example, several
years ago I read that IBM had been developing holographic storage.
Is it still in development or commercialised?

Just a few thoughts.

Cost is always a concern, Even if you have "enough" money, you have to
decide how to spend it.

You can't compare technology that's shipping in mass quantities to
something that's only in the laboratories, in any meaningful way.

magnetic disks are good enough, reliable enough, and the cheapest
solution. When there is market for bigger or faster disk it will
magically become available, cost $200, and the price will drop like a
rock in a few months. That tells me that the manufacturers have
several generations in the lab and are only rolling products out when
many customers perceive a need they are willing to spend money on.
The next wave may be towards 2.5 inch formfactor with the same
capacity (100-300GB) since these are a little cheaper to make, a
little quieter and draw a little less power. Small size is good.
Then there are the postae-stamp size disk for cell phones.

It's hard to compare components and systems that meet enterprise
storage requirements, where price is not the #1 priority with mass
market components where "good enough" and lowest-price rule, although
there is lots of crossbreeding and both sides benifit.

Somebody's Law (I forget who) said that in the long run simiconductor
solid-state always replaced magnetic devices in the marketplace and
predicted the immenant end of rotating magnetic storage. This was when
a 25MB disk drive was the size of a front-loader washing machine, cost
$60k plus $250/month for amaintenance contract, and required a monthly
visit from an engineer for tuneup. They also had a MTBF of about 12
months. There may be a reason why we remember Moore for his law and
don't remember this guy's name. The disk drive engineers have been
saying that they have 2 orders of magnitude (100x) improvement in the
labs. They've been saying this since at least 1970 and hostory has
proven them right. They still say it.

Shock may be an issue for iPod devices but don't see it as an issue
for laptops, or anything else. Battery life in tiny devices may drive
solid state storage here, but it's not going to be cheap for a while.

Lower power is always nice in portable devices but in a laptop
the CPU uses 10x the power the disk does, so there's not going to
be lots of development to make low power disks here. CPUs are
going to be where the long battery life comes from.

The mass market doesn't need higher transfer rates. The current
20-40MB/sec is plenty, even for video. Niche needs like Photoshop can
be met economically with raid0. Ramdisk-like solutions can be
implemented wherever there is a need for zero-latency/high transfer
rates. These will be servers, and it's too specialized to be
mass-market, or cheap.

Any memory made on the same production lines as memory chips is going
to cost $100 to $200 per GB, because that's what it costs to make
chips this way and the market demand supports that price. The newest
postagestamp-sized disks are much cheaper to make and will be in cell
phones ond other consumer devices.

IBM is showing some technology that may replace disk drives,
and I don't think it's a coincidence that IBM sold it's HDD
division to Hitachi. It may happen this time but the demise
of hard disks has been forcast many times.

It's always a mistake to compare the most advanced example of a mature
technology with the first example of a new technology.

 

[J. Clarke]

**** Post for FREE via your newsreader at post.usenet.com ****

Hi all,

I think there are a number of factors such as noise, power,
shock resistance, performance, and heat, that are important
to HDD users.

Forever increasing the RPM cannot really be the answer to
better performance, can it?

Actually, increasing storage density has a significant effect as well.

For a 15k rpm HDD, its access time
is around 6 ms, still about 1000~10000 times slower than
memory-based storage. If we move to solid state instead of
spinning platters, could we then expect frictional heat and
shock to become much less of a concern?

Yes. That is why solid state drives are used in some military and
scientific equipment.

How about using other

methods to suspend the R/W head instead of on air?

Like what? An air bearing is pretty close to frictionless.

A frictionless and/or silent spinning motor design might also be
interesting.

What leads you to believe that the motor is a source of noise in
contemporary drives?

Also, for HDDs, I think:
energy required = energy doing actually work + noise + heat.

If we can remove/reduce the noise and heat, the power required
should also decrease accordingly. In view of current products,
it seems performance and (noise + heat) seem to go hand in hand

Except that drives today offer vastly more performance with much less noise
and heat than drives 10 years ago.

Maybe HDDs are not the best storage device after all? If cost is
not a concern, solid-state drives seem to be superior in every
aspect.

Except that the ones that can hold data across a power failure without
depending on a battery don't seem to be any faster than magnetic storage,
not to mention being good for only a limited number of writes.

What alternatives are available? For example, several
years ago I read that IBM had been developing holographic storage.
Is it still in development or commercialised?

There are all kinds of "alternatives" in the wings. Seems like every time
one comes to market magnetic storage can deliver the same performance for
less cost. Remember when magneto-optical was going to be the Next Big
Thing? Maxed out at something like 5.7 gig on a 5-1/4" disk that cost more
than a 200 gig hard drive? Before that there was bubble memory. Now
there's holographic storage and nanotech storage and organic storage and so
on all coming and what do you want to bet that by the time they hit
magnetic storage will deliver the same performance for less money?

 

[nameruse]

**** Post for FREE via your newsreader at post.usenet.com ****

Actually, increasing storage density has a significant effect as

well.

Is there an upper limit for density?
One elementary particle per bit?

How about using other

Like what? An air bearing is pretty close to frictionless.

But the friction between the air and the spinning platters do
generate considerable amounts of heat (especially at high rpms)?

What leads you to believe that the motor is a source of noise in
contemporary drives?

I suppose only the arm and motor are moving...
Are they other sources of noise?

There are all kinds of "alternatives" in the wings. Seems like every time
one comes to market magnetic storage can deliver the same performance for
less cost. Remember when magneto-optical was going to be the Next Big
Thing? Maxed out at something like 5.7 gig on a 5-1/4" disk that cost more
than a 200 gig hard drive? Before that there was bubble memory. Now
there's holographic storage and nanotech storage and organic storage and so
on all coming and what do you want to bet that by the time they hit
magnetic storage will deliver the same performance for less money?

If cost is not a concern, which ones are the most promising,
technologically speaking?



-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
*** Usenet.com - The #1 Usenet Newsgroup Service on The Planet! ***
http://www.usenet.com
Unlimited Download - 19 Seperate Servers - 90,000 groups - Uncensored
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=

 

[Al Dykes]

**** Post for FREE via your newsreader at post.usenet.com ****


well.

Is there an upper limit for density?
One elementary particle per bit?


But the friction between the air and the spinning platters do
generate considerable amounts of heat (especially at high rpms)?


I suppose only the arm and motor are moving...
Are they other sources of noise?


If cost is not a concern, which ones are the most promising,
technologically speaking?

Go to some web site and see what patents IBM has filed for
data storage in the last 4 years.


It's not clear to me that any alternatives will make a dent in the
desktop PC market, for a couple of years. If IBM (or someone else)
makes some breakthru in storage density it's likely to be rolled out
in some pilot project in some gov't lab that needs massive amounts of
storage, or some sekkrriitt gov't project that has an unlimited
budget. Then it will be a corporate mass storage system.

For an idea what's going on google "petabyte storage". (A petabyte is
1024 terrabtyes).

It's going to be awhile before anything fundamentally new comes out to
replace the disk in your PC.

My $0.02

 

[J. Clarke]

**** Post for FREE via your newsreader at post.usenet.com ****


well.

Is there an upper limit for density?
One elementary particle per bit?


But the friction between the air and the spinning platters do
generate considerable amounts of heat (especially at high rpms)?

So what else would generate less heat?

I suppose only the arm and motor are moving...
Are they other sources of noise?

Most of the noise is the spindle bearings. There's some impulse noise
during seek, that comes about due to the rapid start-stop motion necessary
to get fast seeks, and some turbulence noise, but since turbulence is
badness in the drive capsule for other reasons, they try to minimize that
and it's pretty much inaudible these days.

If cost is not a concern, which ones are the most promising,
technologically speaking?

Don't really know. I used to get excited about that sort of thing. These
days I assume it's all vaporware until somebody actually ships product.

 

[Dorothy Bradbury]

Savvio is U320/SAS, which has a benefit re TCQ:
o Tagged Command Queing optimising seek patterns etc
o Which directly boosts the IO/sec above even WD Raptor

Some SATA drives do support TCQ:
o WD Raptor does now, 2.5" SATA will also shortly
The problem is upstream of the drive itself
o Re controller card, O/S, etc support currently lacking it

A 15k-rpm drive already uses a 2.5" platter - so the move to an actual
2.5" housing isn't so difficult. Savvio isn't particularly thin (22mm if I recall,
compared to 9.5mm for most ATA/SATA 2.5" drives), and so more of
an issue I suspect was the data areal density & move to glass platters.

By 2008 we should have dual 1.8" disks in laptops, good for RAID-1,
and perhaps some freakish designs do 4 for RAID-10. That is already
possible now with 2.5" disks in some widescreen laptops.

So sustained data-transfer is a problem even for current HDs when put
against that of onboard memory. IO/sec is an issue for servers, which is
why for pure server use SCSI still has a home (for now) vs SATA.

18k & 21k rpm SCSI disks are on the horizon, there are limits to the
fluid bearing - mainly re heat/size/density - but as areal density rises,
platter diameter shrinks, so higher rpm becomes less of an issue.

4GB Compact Flash is already commonly available for not huge sums,
basically not so different to 1GB drives in the early 1990s. The problem
is Flash memory has a finite read-write life, and a large chunk of current
MS O/S & Apps use vast numbers of small files regularly rewritten.

RAID can be good for boosting throughput:
o RAID-5 gets faster with more spindles re sequencing of next data read
o RAID-10 gives reliability with speed, WD Raptors are fast for SDTR

RAID however can't speed up some operations by much:
o Lots of small files are the enemy
---- most of the time-to-transfer isn't in the bit density
---- it's not so much in the actual rotation of the platter itself
---- it's in the movement of the head itself, the seek latency
o Shifting a huge number of tiny files can cripple "headline" SDTR figures
---- particularly true where cluster size exceeds a single file in size
---- thus RAID offers little benefit, seek becomes the bottleneck

Unfortunately we are creating O/S & Apps, Roaming Profiles and lots of
similar stuff which creates a many-but-tiny-file culture to the data-set. I do
wonder if MS-IE is designed to grind machines to a crawl, same with OE
when a colleague set every individual person's email to their own folder.
Drives the backup s/w & Anti-virus s/w mad too - on a laptop, hilarious.

I'm not too convinced about the efficiency of XP re caching apps/data.
It may be a function of the bloat of much of the Windows s/w & tiny files.

Data-density will continue to increase, I think the projection is 400GB on
a 2.5" disk by 2007 and 1TB from a 3.5" some time before that. Wonder
how long a RAID-1 array rebuild of 1TB of data would take, oh gee...

The industry does seem to be pushing for 2.5" in RAID form from both a
thermal (airflow thro 1U limited by c/sectional area left after a few 3.5" HD),
and an availability viewpoint. RAID1 isn't a backup, but boosts reliability.
Some cute 2.5" desktop & server tiny RAID boxes for SATA 2.5", the
scsi 2.5" are a little chunkier which limits their density a bit more.

Soft-edging the seek helps reduce noise, but frankly even the so called
silent laptop drives when thrashing away make a fair amount of seek noise.
Seek time is important for server (v workstation) disks, although watching
a workstation thrash around doing an anti-virus scan it matters there too.

The major heat factor in HD is the fluid bearing/motor assembly. You do
not want the oil viscosity changing significantly, which excess heat can do.
The 15.3k drives use fluid, some 10k-rpm drives still use ball-bearing (the
early WD Raptor did too), so note which a noise sample is using. Cheetah
drives doing a I/O seek thrash sound like frozen peas poured onto the base
of an aluminium pan - the idle dB(A) figures are really moot in that respect.

Wattage has fallen considerably per unit performance, early SCSI drives
were toasty without considerable cooling - now it's not particularly hard to
cool an array of 18 SCSI drives thrashing away in a rack enclosure. Some
workstation cases have their intake port a) too small and b) beneath the
area of the drive cage so denying drives even basic active cooling.

Silicon storage would most help servers with lots of tiny files re IO/sec,
many technologies on the horizon and disk isn't quite dead yet. Indeed,
near-line storage with cheap SATA drives is a strong growth segment,
offering vastly higher backup speeds, economically, than tape systems.

Physical redundancy of media favours disk, since alternatives are likely
to be more expensive. So disk will be around for quite some time.

 

[nameruse]

**** Post for FREE via your newsreader at post.usenet.com ****

So what else would generate less heat?

This is the reason why I was thinking if there are any
methods to suspend the R/W head without relying on air.
Then vacuum can be achieved inside HDDs which should
reduce the heat production due to friction. I would
think the construction of modern HDDs are sturdy enough
to withstand the pressure difference.

Most of the noise is the spindle bearings. There's some impulse noise
during seek, that comes about due to the rapid start-stop motion necessary
to get fast seeks, and some turbulence noise, but since turbulence is
badness in the drive capsule for other reasons, they try to minimize that
and it's pretty much inaudible these days.

If seeking noise is made by the seeking actuator, does
lubrication help ?
IIRC, sound waves cannot propagate across vacuum.
I wonder if it is possible to make a double casing for
HDDs; these casings are held in place by force
(magnetic?) and the space in between would be near vacuum...



-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
*** Usenet.com - The #1 Usenet Newsgroup Service on The Planet! ***
http://www.usenet.com
Unlimited Download - 19 Seperate Servers - 90,000 groups - Uncensored
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=

 

[J. Clarke]

**** Post for FREE via your newsreader at post.usenet.com ****



This is the reason why I was thinking if there are any
methods to suspend the R/W head without relying on air.

You can suspend things using an air bearing, a liquid bearing, an
electromagnetic field, a very, very rigid structure, or what else? A
liquid bearing gives the same disadvantages as an air bearing and adds a
whole bunch of its own. An electromagnetic field is out for obvious
reasons. That leaves a very, very rigid structure, which means lots of
mass which means lots of inertia which means lots of power required to move
the heads which means a large increase in seek noise. And since seek noise
already dominates, you haven't made things worse instead of better.

Then vacuum can be achieved inside HDDs which should
reduce the heat production due to friction. I would
think the construction of modern HDDs are sturdy enough
to withstand the pressure difference.

Sorry, but that is not the case. There are some disks made with sealed
capsules for use in very harsh environments (military, space, etc) and
those disks have considerably heavier structure than current consumer
drives.

If seeking noise is made by the seeking actuator, does
lubrication help ?

No. It's the shock of starting and stopping the drives that generates the
noise.

IIRC, sound waves cannot propagate across vacuum.
I wonder if it is possible to make a double casing for
HDDs; these casings are held in place by force
(magnetic?) and the space in between would be near vacuum...

You could do that, but now you have your drive encased in a Thermos bottle
so where does the heat go?

 

[nameruse]

**** Post for FREE via your newsreader at post.usenet.com ****

You can suspend things using an air bearing, a liquid bearing, an
electromagnetic field, a very, very rigid structure, or what else? A
liquid bearing gives the same disadvantages as an air bearing and adds a
whole bunch of its own. An electromagnetic field is out for obvious
reasons. That leaves a very, very rigid structure, which means lots of
mass which means lots of inertia which means lots of power required to move
the heads which means a large increase in seek noise.

I don't know how electric field or magnetic field would work
to suspend the R/W head, but I'd imagine it could be done,
depending on how data is read/written? As mentioned, a rigid
structure is probably heavier and requires more power, but an
increase in power usage doesn't necessarily translate to added
noise, does it? If it's the start/stop shocks that produce
noise, it could probably be remedied as discussed below.
Besides, anti-gravitational force might be of use.

There are some disks made with sealed capsules for use in very
harsh environments (military, space, etc) and those disks have
considerably heavier structure than current consumer drives.

If we can't make it to withstand the pressure without making
it big and massive, then it's inconvenient and probably a
good idea to wait for stronger and lighter materials.

You could do that, but now you have your drive encased in a Thermos bottle
so where does the heat go?

It's integral to consider the heat issue.
Indeed, the rate of heat transfer is much slower in this case.
But if we could reduce the heat production in the first place,
as discussed above, this should be less of an issue and would
virtually eliminate seek noise and any sound generated within.



-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
*** Usenet.com - The #1 Usenet Newsgroup Service on The Planet! ***
http://www.usenet.com
Unlimited Download - 19 Seperate Servers - 90,000 groups - Uncensored
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=

 

[John Turco]

nameruse wrote:


No. It's the shock of starting and stopping the drives that generates the
noise.

--John
Reply to jclarke at ae tee tee global dot net
(was jclarke at eye bee em dot net)

Hello, John:

I have a pair of Samsung SP1614N (160GB PATA) hard disks, and they're
absolutely silent. Samsung boasts of its own technology (NoiseGuardTM,
SilentSeekTM), used to make its drives so quiet.

The company supports the industry-standard, "Automatic Acoustic
Management" (AAM), in addition. Further, it claims the following:

"While AAM feature provides the best seek acoustic noise reduction, it
can also cause benchmark performance degradation of the drive
curtailed by the elongated seek. According to SAMSUNG's evaluation,
the benchmark data transfer performance in quietest AAM mode is
degraded approximately by 3% - 5% from that of a drive using only
SilentSeekTM technology, while overall seek acoustics is reduced by
only 0.05 bel - 0.1 bel (average sound power). Therefore, the benefits
of SAMSUNG's SilentSeekTM technology can virtually outshine those of
AAM feature."

[Quoted from: SAMSUNG Digital World - Hard Disk Drive/White Paper
http://www.samsung.com/Products/HardDiskDrive/whitepapers/WhitePaper_03.htm]

Bold assertions, perhaps, but I'm a satisfied Samsung customer!


Cordially,
John Turco <[email protected]>