Just a though - HDD design

This idea came to me when I was looking at the HDD & SSD thread.

Now, HDDs - as far as I know there is a "set of arms" on one side to read an write data from a drive.
The physical repositioning of the arm overal lowers performance.

Now I just thought, what would happen if HDD manufacturers installed another "set of arms" on the opposite side of the HDD.

I don't think it has been done before, or has it?

And if, youldn't this signifiantly increase read/write speeds?

 

This would be cool, but the drives would have increased power consumption, and would probably be twice as susceptible to head-crashes. :0

Also, if you add about another 62 arms and stop the platter spinning, you get SSDs.

I think they did something like this in Multiread CD drives. <- wrong name....

 

Wont work, well at least in a 2.5" drive.
The arm you are referring to is known as an actuator arm. It is controlled by a small microcontroller inside the hard drive. The actuator arm floats in a magnetic field which allows the head to move back and forth so fast. The actual size of this is rather large, it takes up basically the rest of the drive.
Here is an actual image of the inside of a hard drive,

This drive is dead, but I pulled up the actual head on the drive, so you can see it.
The whole part next to my finger is the whole arm control mechanism. To the right of that may look open, but that's where the microcontroller and maybe small PCB inside the hard drive is.

It may be more possible on a desktop hard drive, but not in a laptop one.

Their is also another thing you should realize. It doesn't matter how many heads or arms their are. If the hard drive manufacturers wanted to, they could put 4 actuator arms, and quadruple the number of heads in the drive. But this doesn't increase performance. Why?
This is why
When a hard drive goes to get data, it can only fetch one piece at a time. Think of it as a two way street. The drive can get data or send data to the drive. The head is controlled by the microcontrollers on the reverse side of the hard drive. They pool the information together, and organize it for efficiency. Than one by one each packet is sent or retrieved from the hard drive.
If the hard drive has 4 heads, that means it covers 4 sides of two platters. However the drive can only read/write from one at a time. Thus having an additional head is a waste. Its only done to increase data capacity, for multiple disc drives. If this were untrue, a 160gb 7200rpm drive would be half the speed of a 320gb dual platter 7200rpm drive. This is not true, they are both very similar in speed, with actually the single platter drive being slightly faster. This is only because the data needs to be split up to two heads, rather than 4. Splitting the data in half is easier, and thus yields slightly faster seek times.

K-TRON

 

Would increase the size of the drive as well.

EDIT: Wow, my post looks like nothing compared to that above me XD

 

I didn't get all of what you said K-Tron - possibly because my mind is sort of semi switched off.

Basically you said its not possible - OK - too bad

Thank you for the detailed answer

+ Rep for that

 

Basically think of it this way, even if you have two arms that are on opposite sides of the platter, when one needs to read data the other one can only read data that is on the exact opposite of the platter, i.e. when head 1 is reading a red section head 2 can only read the opposite red section and the chances of the data needed being in that section is slim to none. Now if the arms could moved in 2 dimensions then it would work, but that would negate the need for platters to spin at all.

 

Ummmmm.
You're confused on how a hard drive works.
That arm does move in two dimensions. The drive spins so it can read all surfaces of the disk, but the arm moves as the head is only a very very tiny portion of the whole disk surface. I will attempt to illustrate, then edit my post.

EDIT: Here we go.

This is how you're thinking a hard drive works, but that would limit the usable platter space to what is highlighted in red. The read head is only a very small tip of the read arm.


Instead, it's more like this. That arm is on a motorized pivot point that rapidly moves the arm back and forth. This range of motion allows the tiny head to be able cover the entire surface of the drive



Hope this helps.

 

K-TRON
Hep
jakejm79

hats off to u guys, this forum is very educational indeed.....
keep the knowledge rolling.

 

\No I understand what you are saying, maybe my illustration was too basic,
Stealing you illustration I will try to show my point
Basically if the data you want sits anywhere on the green lines of movement, then you will notice that that well head 1 is reading any part of the 1st green section head 2 has to read anywhere on the yellow section, and since the additional data that is on the other green line line that is a section that the 2nd head cant reach, in fact if the disk were spinning counter clockwise, it would pass head 1 before reaching head 2. My point was the arm/head only has 1 dimension is come by the platter spinning.

 

you could read from different positions though... or not?

 

You can, that's my point.

 

The overhead of doing this would be emence and likely provide a lower output in the end.

 

I don't agree with some posts posted above, especially with K-Tron's post.
Just imagine you have a HDD ( A ) with one platter and one arm, with 2 heads to access both sides of the platter.
Now we have a second HDD ( B ) with two arms, each arm has two heads and they are on their opposite side.
I know this will increase the HDD size, but that's not important at the moment.

Now I have 10 files on my HDD which are placed across the HDD.
An old HDD will read file after file, starting with file 1, ending with file 10, and the platters need to rotate, lets say, 8 times.
A newer HDD ( A ) will optimize the reading, using the cache and several techniqes like NCQ, and the platters only need to rotate, lets say, 4 times.
HDD ( B ) now, uses an advanced NCQ technique which is optimized for two arms, and thus the platters only need to rotate maybe 1 or 2 times, collected in the cache and send in the right order to the chipset.

So a HDD with two arms will be faster, don't know why it shouldn't work, it will be a lot faster, especially regarding access times.
You can also say, that you have a HDD with two arms whereas the first arm only accesses the first half of the platters, the second arm the rest. This will also increase the seek times, because the arms have to travel across a shorter distance.

So it works, it's just difficult and too expensive.
You have to add everything twice, except of the platters. This increases the risk, because now you have two arms which can fail and it also increases the necessary processing power and you need better algorithmss to manage two arms. You'll also need more space for the arms, but this means less space for the platters, which results in less capacity.
So it's just too difficult and it will become too expensive.

 

Different positions on that platter? Basically the head travels back and forth over the radius of the platter so at any one instance in time it and read any data on that radius but because each head as a pretermined range of motion (the arm only swings back and forth [it doesnt move up and down or side to side]) the 2nd arm would only be capable of reading what was on it's range of motion radius at that time too, true you could get lucky and have the additional data you want there or it could be quicker to let the data reach that head rather than the other head, but the increase wouldn't be significant and frequent enough for the increase in cost and engineering to produce it. Maybe something like this would be a little better, Head 1 acts like a regular head, while head 2 specifically handles the pagefile and/or MFT duties therefor reducing the range of motion (and there for time) that the heads can travel, but the data is separate enough that it may provide significant benefit. ::: Sorry if my wording is poor, I am having a hard time translating my thoughts :::

 

Get outta my head

 

FrankTabletuser, It ha been proven that the number of heads does not make any difference on performance. If your statement was true, why do single platter and double platter drives perform roughly the same.
Your statement on NCQ, and optimized organization is spot on. However the drive heads only read/write one at a time. So even if you had 30 heads, versus 4 heads, they would perform pretty much the same.

Having two actuator arms is controversial. The way the question was posed, was two arms on opposing ends of the drive. I do not think that will work any different than a normal hard drive. It would require basically 2x everything including the PCB/microcontroller board. Doing this would be a bad idea. It would double power consumption and increase the heat generated by the drive. If it was a good idea, I am sure the hard drive manufacturers would have come up with it. But they didnt.
The only way that a two arm system would be any better was if the arms controlled independent parts of the disc, like jakejm79's image in post #14. That may actually work, but I have my doubts on it.
The only way to increase hard drive speed is to redesign the way in which files are stored on the drive, and the magnetic sensors inside them, increase spindle speed and increase data density.

K-TRON

 

this is incorrect IMO. The arm can not align (in a line) from it's center base point to the center of the disk, because this way the head has to turn 90 degrees in order to read/write.

instead, it touches (tangential) to the center circle of the disk.

 

anywayzzz, a real improvement of HDDs would be if they stop using arm and a moving head-style design.

They should make one non-movable arm from the side to the center of the disk with tons of heads on it, one for each track. This way there would be no access time and they would be able to compete with SSDs

 

1. We talk about two arms, not only a few heads more, a whole arm more. So every comparison with a single and double platter drive is nonsense.
2. When we use two arms (yes, then we're also forced to use twice as much heads) then we can access both arms separately. This means that both arms can access a different location and different file at the same time and can read this different files at the same time! (this is not possible with a two platter drive, because all heads are at the same position, just on a different platter)
3. Such a thing will work and it will increase the performance because they've already "build" something similar, also called RAID 0, and maybe that's why they don't build such a drive, because then you can also create a RAID volume and you'll get something similar, and it's cheaper to create a RAID volume than to create such a specialized hard disk.

They could also split the single arm in two arms, if you have a two platter drive, which are still located at the same position, so that the needed space stays the same. The only problem is that it may get difficult to control them with the coil if they are located at the same position.

I don't think that it will increase the power consumption that much, because you still have only one motor for the platters and you also don't have to double the whole PCB board, maybe only add 25% more parts. It's maybe even possible to reduce power consumption because you can switch between different modes. In one mode the two arms move like normal, in a power saving mode their movement could be reduced (they only access a part of the platter, like jakejm79 picture shows) and thus power saved, because they don't have to move across a large distance, and thus the current for the coils which move the arms could be reduced.

I don't know why they don't build such a drive, but I think they've had such drives in development, just canceled it, because it was too expensive, they've failed too often and RAID 0 does something similar.

 

Oh I see what youre saying. Yeah I dont know how it would work. It either doesnt do anything or maybe increases performance a little bit.
If you were to replace one head with two, problems would start happening. I say this because with one head, the head is directly above the data. If their are two side by side, they cover different circumferences of the disc, and that would be ridiculous to figure out how to control it.
I do not know how the harddrive actually fetches data from the drive. Knowing that would probably help.

K-TRON

 

Reading with two heads would only be a problem if you wrote with one head. If heads wrote two pieces of concurrent data, it would be read that way as well. So no problem

 

The only problem is the density at which the heads would have to be lined up next to each other would be incredible, you would have to sacrifice data density/capacity and increase cost and at which point the money would be better spent on a SSD. Problem is once you stop the head(s) from moving then they can only read/write on a particular circumference of the platter. In theory the ideal answer would be to cover the platter with heads, therefor eliminating the need for the platter to rotate or heads to move, but that would suffer from the problems above even more so.

 

^ well they don't have to make it with one line of heads due to the head size being bigger than the track that it will read/write. They can do multiple lines with heads, so that the density at one line doesn't have to be that tight. Say the first line with head will cover tracks 1,5,9, ... and the other one will be on tracks 2,6,10, ... etc.

the actual head is no big deal to make. It's like a small coil .. that's it. The work however will lay down to the development of a new style controller.

and the disk has to rotate no matter what, since it would be one head per track.

this is the only way access time of the HDD can compete with the access time of the SSD.

 

Ahem...yes, it has actually been discussed and done before. -> http://www.storagereview.com/guide/actMultiple.html

Apparently the main problem is additional cost for relatively low gain.

edit: and here's a patent http://www.patentstorm.us/patents/5343347/claims.html

 

OK... this settles it then I think

 

i know that raid1 does this on some controllers for read (well, behave similarly at least).

this is great in servers, where a lot of data has to be read simultaneously.

one could "force" the different heads to be one at the outer half, one at the inner half, for lower max latency.

for an ultra high performing disk this could be done. but imagine it to be a 5.25" disk which would have to be buffered for the now big chance of a head crash. raids can provide similar behaviour (for read only, though) with additional redundancy.

i'd like to have multiple lasers in dvd/blueray drives. for fast reading of them, would be great

espencially in games, where there has to be quite random access, they wouldn't need to dublicate game data all over the disk for fast access, anymore, then (ps3 is terrible at doubling, tribling etc the data amount just for fast load time).

could be sort of the highend ps3 or xbox would have more read-lasers.