Better Hard Drives??

[chewy0914]

I looked at hard drives in the light of what parts of the pc slows it down.

What are the limits of disc based hard drives?

I found spin speed? Arm movement? maybe circuitry?
Size of the disk?

I read a few post on how it has been tried to make them faster.
Is it just cost or what is limiting hard drive manufacturers?

Thoughts--

2 arms in a hard drive? both able to read and write at the same time?

Multiple arms so any part of the hard drive can be read from in the same amount of time?

Disc- Rotated from the outside keeping the center of the disc?

Magnetic rotation instead of motor based? mag lev stabilization?
Spinning and holding from outside/middle/inside of disc?
Or even roller/other based?

different kind of arms?
or no arm at all?
Arm that allows each layer of arm to be in whatever position from tension or other?
1 arm motor moves layer of head each head in set ways?
*

Strip/other going the whole length of hard drive able to read or write from entire disc surface at once?
*

*multiple arms/strips that do do don't move to cover entire area.

Better aerodynamics inside of hard drive reduce friction?
Not just helium-- internal structure.
Increase speed from air flow produced?

Better heat disbursement?
Air flow in around hard drive?

Heat Sink?

Sealed area around disc..
Air flow around area in harddrive..
Reverse flow blow out dust?


Single surface that covers entire disc?
Be able to read or write from any part of disc at once?
Rotating or not?
Example:
Cones multiple read heads per cone/ mechanical/other redirect
for multiple locations?

head head
\ / \ _ /
\ / \_!_/switch
disc disc


Raid hard drives:
Can this already be done?
Half+ data on different parts of disc?
4x+/- with say 2TB drive with data placed along the disc repeated 4 times?
making 4x512gb with 4 spots to read from? 3 backups?


Also what is the limit in write speeds?
Multiple discs and heads on 1 arm?
Write each bit of data one after another on different disc?
Then put them together or however in down time?

[HaQue]

There are already engineers working on this at levels that would blow your mind.

I have read through some papers that go into so much detail, it is obviously work from an entire professional carreer.. many of which the people involved have been at universities for 20 years - and are still there.

For a performance point of view, the best "we" can hope for is to research each part of the system on what is available and what is good in terms of performance, but not forget the application you are using it for. A "games" PC with excellent performance might not be any good for a high performance file server for example.

Also, you have to have your limit on what is "good enough" otherwise you spend all your time researching and are never happy.

Growing up in the LAN Party world, I had average PC's but mates of mine would hold out buying better stuff, and when they did finally buy their kick-ass GFX card or whatever, had issues with drivers. RESULT: I spent most of my time playing games, they spent most of their time frustrated.

For you: google terms like "whitepaper" "hard disk performance" etc and you should get some answers.

[digitalferret]

Wild guess but what prompted this thought process? Is this a college assignment?

If so maybe the point is "thinking outside the box".

Don't you think maybe you can only refine a fundamental design so much before some returns are so diminished as to make the "upgrade" untenable.

Doesn't it take a quantum jump in order to make a significant improvement in speed,
ie Aircraft speed: pistons versus jet engine. A new technology.
We still use piston aircraft, just not in the pursuit of high speed.

Isn't the main driver behind improved hard drive design the balancing of high speed with useful capacity at an affordable cost?
Don't you think the main advantage of spinning disks is affordable capacity increase rather than speed of operation?

Just a few thoughts to ruminate on.
Good post tho, gets the cogs turning ... or should that be synapses firing.

[HaQue]

"Strip/other going the whole length of hard drive able to read or write from entire disc surface at once?"

This is actually an interesting idea, though I have NFI how it could be implemented, if you could read from anywhere along the "bar" at any time, you could do away with the traditional moving heads. would only need 1/2 a rotation to come back under a "head". I am sure if it was viable it would have been engineered long time ago.

This is what happens when you are eating and cant touch the PC, and get too much down time to think...

I probably should shutup before you guys think I am also on crack!!

[Keatah]

Singer/Librascope magnetic memories had 1 head per track.
http://www.dvq.com/oldcomp/misc.htm

Attachments

[HaQue]

An inventory control system on 1/2 a megabit... wow

startup time 5 minutes, thermal stabilization 2 hours. holy Toledo.

Thanks Keatah, very cool

[Keatah]

I would like to add that many ideas of the 60's and 70's were not practical due to manufacturing issues, like process size or computational power. They may be doable today, but still practical because of cost.

The number one speed limiting factor in spinner disks is that they are mechanical. It takes time to move things from there to there. I personally believe that spinner disks will hit a plateau in performance/capacity in the next several years. They may be relegated to offline storage and archiving. I believe that the new V-NAND (look it up) and similar solid-state technologies will supplant HDD.

SSD already rules the nest in portable devices like newer laptops, all phones, and all tablets. And that's a significant share of the market. SSD is poised to advance in capacity this year. We're already at 512GB and 1TB. They are almost cost effective too. Plus all their added benefits. It's win-win.

Remember that holographic crystal storage kick they were clamoring about in the 1980's? Well it's here already. It's the traditional SSD. Sort of crept up on you. Silicon crystals.. All that is SSD. It's here, and now instead of the old planar stuff it's gone in the Z-direction. Building upward! And accessing these vertical layers is stupid simple too.

But, anyway, let us see some of the tech you propose and why it is or isn't in use.

Multiple actuator arms:
IIRC Quantum experimented with a disk with multiple servo actuators, multiple sets of arms moving independently. They found the complexity outweighed the speed increase. Maybe today it would be practical with the better microcontrollers available. But then you have cost of a second mechanism..

Individual arm movements:
Again not really. Cost. And you can parallel read/write just fine if you move them as a whole solid assembly. You use the electronics to control which head(s) are active. A little bit of software and hardware cache is far more effective.

Strip heads:
It would be costly to manufacture a thin strip head running the radius of the disk. The yield would be low and then how would you deal with spindle wobble and fine tracking? Furthermore you'd have to make this strip-of-heads have gaps. And the tracks would need to be wider to deal with temperature expansions - which is non-linear across the surface.

Magnetically levitated disc:
You could not achieve platter stability if you used a linear motor "pulling" and suspending and spinning the disk in mid-air, not in a commercial environment. It may work for magnetic toys and science experiments and the Mag-Lev trains, but not for disk drives.

You can achieve the necessary precision as demonstrated with Gravity Probe b and its near perfect spheres. But not in a cost sensitive mass market. http://en.wikipedia.org/wiki/Gravity_Pr ... ntal_setup

Roller based rotation:
Noise, vibration, wear.

Multiple stationary arms covering the whole area:
Way too many tracks and temperature expansion issues. This idea did work in the 60's and 70's as shown in the pictures above. But note the warm-up and stabilization times. And the low data density.

Spin speed:
Heat generation, friction, and power consumption are issues. It is possible to build a 20K RPM disk, but power and heat make it less practical.

Read and write at the same time:
Precision timing, power, and local interference of magnetic fields make this currently impractical.

Raid hard drives:
Cost and firmware customization is my guess why this hasn't been done. I see no advantage here over conventional multi-HDD contemporary RAID setups. And this would be as prone to failure as any single disk. No reliability improvements. In fact, it could be worse due to the complexity.

Now that you know why your ideas aren't in use, keep thinking! You may come up with something to change the whole industry yet!

[Keatah]

The whole issue with consumer systems being bogged down by the mechanical HDD has to do with modularity and paging of the operating system. Loading and unloading modules based on what’s needed. This was a concept invented in the 1950’s out of necessity. Invented in a time when electronic storage was insanely expensive. Thousands of dollars per kilobyte! The early disk drives reduced that cost to hundreds of dollars.

And an engineering team at IBM came up with the idea of having the processor only keep what was necessary in memory to do the immediate task. Even variables were offloaded to rotating storage devices. Which is what the Singer/Librascope memories were. They weren’t disk drives, they were mechanical RAM.

At anyrate filesystem paging and loading and off-loading of modules was born in the 1950’s and has been with us ever since. Operating systems may continue to be written this way for the foreseeable future but the SSD masks a lot of those delays and will eventually get to primary storage speeds. It’s already there in some high power applications. And half-way there with PCIe Flashdrives.

It is my opinion the modular and paged operating system is a complete bastardization and seriously outmoded philosophy. Furthermore, mechanical HDD should only be used to store archival information and working data sets and NOT ACT AS SYSTEM RAM because programmers are too lazy. Sorry, but you kow that’s the truth. Like it or not.

And mobile computing devices are going that way. You can hook them to mechanical disks, but you won’t find mechanical disks inside them anymore aside from the cheap laptops. And in that context, the mechanical disk becomes an archive medium. And one that needs to be protected and coddled and backed up, still.