if only they could make small enough HDD heads

just thinking, if they could cover the disk on both sides with heads - one for each track ... then access time will be insignificant, no chance of arm controller fail, no chance of head hitting the disk, less heat, etc ... and those drives will be able to compete with the SSDs. It would be only one spinning disk covered with heads on both sides.

I mean what would that take ... like some 200,000 heads by side for a drive of like 100GB

 

Won't work lol.

If there is one head and one head on the other, that means the head only has certain areas the head can write in, while the other has to wait it's turn for the disk to rotate and put the 2nd head in the correct location. Vice versa.

It would actually increase access times because if the 2nd head needs to write to the other side of the disk where the 1st head is, that means it will have to wait for the platter to spin, meanwhile the 1st head is staying still, nothing doing anything.

Now if you say well you can write algorithms to account for that, yes you can however you do realize how scattered information would be on the platter. You would have bits and pieces everywhere, and that will lead to uberly horrendous read/write times, as well as increased access tiem.

 

I think you are confused. There is already a separate head for each side of each platter. What the OP means is like a million heads all in a line on each side of every platter.

 

I think I remember a drive (Seagate?) where it had four heads on each platter - but it was not very good (or at least no one wanted to perfect it).

Trottel - I agree.

 

Okay I think this was very badly written. I think I kind of understand what he means, now. Anyways those would have be some incredibly small heads.

 

there are plans for it, but without spinning platter.

with the spinning platter, it would massively increase the head-crash chances + each head would consume power if they're movable in some way.

it could work if done well (NCQ would help there massively), but it would be costy, and have high mechanical risk. i prefer ssds.

 

If the platter isn't spinning, the heads will have to, right?

The heads are already small enough - the surrounding electronics (to position them) is what makes them 'huge'.

 

Eh the head's themselves are small, however the arms are not.

Also wouldn't this be ridiculous to design? I mean imagine trying to hit say 200,000 heads, writing/reading information simultaneousness. We would very literally need to design a brand spanking new I/O system (Hard Drive Express Lane by 100x pins?, perhaps optic fiber would be necessary?

 

Yeah, a single, moving arm is not small - but a fixed arm that doesn't move can hold all heads easily (they may not be in the same line, but the head size nor the arm size is the limiting factor here).

No new I/O system needed either - just like the Hybrid XT is invisible to the O/S, the fixed, multi-head drive's internals can be kept 'secret' too.

Basically, the onboard controller would be handling the new physical head arrangement - just like SSD's do too.

 

Actually, hard disk tracks-per-inch have always been far less than bits-per-inch because of the need to mechanically position the heads to change tracks. Even with servo data in the track, it is still only an approximation of the correct location and the drives need to leave "guard bands" of empty space between tracks.

As a modern example, take a look at the Seagate Constellation.2 SAS Product Manual (caution - big PDF). The drive has 270,000 tracks per inch, but has 1,544,000 bits per inch - nearly 6 times the density.

A major drive manufacturer (I can't say which) experimented with different types of head technology in the 90's. IBM was using what were essentially integrated circuit wafers as disk heads, and the other manufacturer decided to try to make a "linear head" drive, where the "head" was actually a fixed row of thousands of heads using a layout similar to to the repetitive patterns found in memory chips.

They got as far as the prototyping stage. But alternative technologies* proved easier to implement, so they didn't pursue it.

* Like perpendicular recording. Despite the first PMR hard drives being delivered in 2005, the technology is a lot older than that. I had Vertimag floppy drives in the 1980's that held 5X the capacity of conventional drives, for example.

 

Terry, thanks for expanding on what I (tried) to say.

Instead of 'surrounding electronics', I should have said 'surrounding electronics and mechanical considerations'.

But you still stated it more eloquently than I did.

 

The biggest problem I would see is that I'm not sure how cost effective it would be compared to a SSD. Sure, since we're back to platter technology, data density would be quite high, but the question then becomes how expensive it would be for all of those heads, as well as the controller necessary to coordinate them all. As well, this doesn't solve the head crash issue; as long as you have a spinning platter, you have the risk of a head crash. Heat won't really be less either, in fact it might be more (more surface are of the platter covered, retaining more heat). I'm not saying this might not be worth pursuing, but I think at best it would be a bridge between current HDD technology and SSD technology; just as fragile as HDDs, but with speed closer to SSDs.

 

Kennedy you are forgetting some stuff here. The reason why there are more bits than tracks is not due to guards but due to physics.

a track runs across 1 inch of surface in a linear space and the bits are points in that line

_____________________
..................................
_____________________

That is a single track with bits. there is more bits due to bits are points in the track. It has nothing to do with guard space. I hope that made sense i am bad at explaining. Never knew any of this before lol used logic and came up with that ^^

 

think about it....why would a track of data only have 1 bit per inch? does that make sence? Of course not. A track will have multiple bits per track so my "diagram" makes complete sence. Also this guard space does not mean there is less bits per track but less tracks therefor less bits.

not to scale* lol
| one iche|
______________________
...................................
______________________
XXXXXXXXXXXXXXXXXXXX
______________________
...................................
______________________
XXXXXXXXXXXXXXXXXXXX
_______________________
.....................................
_______________________

see what i am saying? Now if you through in another track in the dead/guard space than you would add another track with X number of bits but it wouldn't be 1 bit per track or 2 tracks per bit. That is simply illogical as Mr spok would say ^^ or as his father would say illogical illogical ILLOGICAL ILLOGICAL!!!!! hahha i loved that part


granted this may be completely wrong if i am miss understanding what a track is ^^
http://en.wikipedia.org/wiki/Track_(disk_drive)
checked wikipedia and i think i am right

 

no, moving. not rotating.

it's called millipede memory

Millipede Memory

 

Unlikely, as I used to be a disk drive designer

I'm not quite sure what you're getting at, so let me try again...

bits-per-inch are the bits laid down in a circle as the drive spins. This used to be limited by how fast the electronics could respond and/or the size of the magnetic area that the the head magnetized as it wrote. These days it is limited by how closely spaced the head can create these magnetic areas without the areas affecting each other (magnetized areas attract or repel, just like whole magnets). The recent PMR increases density by stacking the magnetized areas vertically instead of horizontally as on older drives.

tracks-per-inch, on the other hand, select which of the above circles are being read/written by the head. This involves moving a relatively massive chunk of metal and plastic (the head actuator arm) across the disk surface to find the correct circle (track) to read/write. Because this is a mechanical movement, it has much coarser spacing than the bits around the track (circle). This is due to a number of things, including repeatability (how accurately it can get to the exact same area) and jitter (think of jamming on your car's brakes to come to a fast stop - your car bounces around as it stops). The heads do the same thing - you want to get them to the desired track as fast as possible ("seek time"), but to do that you need to stop them very fast, which causes vibration which reduces positioning accuracy (unless you want to wait, doing nothing, until they stop vibrating).

Perhaps a couple of pictures will be useful. This is the head assembly from a relatively modern 3.5" disk drive - a 75GB Ultrastar:



That whole assembly pivots to change the track the heads are over. The loops on the right are the "voice coil" which is pushed and pulled (just like in your stereo's speakers) to pivot the heads on the center bearing.



On the left side you can see the "head stack". On an actual drive, there would be a disc platter between each paired set of heads.

 

It would be quite cost-effective. It uses most of the components from a regular disk drive, but gets rid of all of the head positioning mechanical stuff and related electronics.

As I described in my prior post, the heads were built using a similar process as a memory chip, and just like a memory chip it is easy to select which bit (head) to read or write.

This would have been a modern implementation of an old concept called a "fixed head disk" - in fact, the earliest systems (disk and drum) normally used fixed heads.

 

Well, if it were a straight arm across the disk so each "groove" of the hard disk could be read at any given time, the only access time would be the time for the disk to spin the given angle to reach the heads. It would probably require a different way of storing data, almost writing in "parallel" so it could read from each track simultaneously. Would take a significant design change but could be workable. Have the arm span the entire diameter of the disk to make it even better.

Lol. I have no idea what I'm talking about.

 

The question is, though, how expensive would it be to make all those heads? I guess a better question is, what's the most expensive part of a HDD? For a "regular" HDD, I would guess the platters, but if the 4 heads (for a twin platter drive) cost, say, half as much as both platters, then the fact that we're significantly boosting the numbers of reading heads would raise the cost by a lot. That's really more what I meant by cost effective; would the rise in price for the components of this format be cheap enough to make it worth pursuing.

 

you totally missed what i was saying

 

I think it would be quite an improvement over the regular HDD that we see today, and will even offer benefits over the SSDs, such as retaining data for much longer time when the unit is not in use, unlike with SSDs that would eventually leak the charge out one day.

back to the point - there is no need to have only one line with heads. This will only make it difficult to fit many heads on one line. Instead, there could be multiple lines all offset by the number of tracks needed.

Example with 10 lines (36 deg apart from each other) with X number heads per line:

- line 1 reads tracks:1, 11, 21, 31, 41, etc ... with head number 1, 2, 3, 4, etc. (all next to each other)
- line 2 reads tracks: 2, 12, 22, 32, 42, etc ...
...
- line 10 reads tracks: 10, 20, 30, 40, 50, etc ... same thing.

^ this could significantly improve the density of tracks on the disk, so we can have with times the capacity of the todays HDDs while using same disk materials. On a normal HDD today there's space left between the tracks to create margin of error for the positioning arm, as it may not hit the exact same place of where the track is.

this will of course require a totally new controller to handle the way this would all work, but I dont think it would be that hard to make. There will just be a bunch of connections from all heads coming to couple multiplexing chips that can address the heads as needed. Then the controller could activate the needed head and then use data link to read/write with that head. That's about it.

Later models could use couple heads at the same time, thus significantly improving the performance of the drive, much like todays SSDs. Think about it - using 10 heads at the same time (like one from each line in the above example) will increase the speed 10 times ...

- the more lines with head one can fit, the better performance/size could be achieved. Hell why even lines - there could be a disk (not moving) all perforated with heads sitting on top (and bottom) of the data disk (that is in the middle) This way all the connections to the heads could be cut on that disk much like on a chip today, and the multiplexing chips would be placed on that disk as well.

same concept, one head per track

 

hard drives loose data too. there magnetism drops over time

 

Yes but there are bandwidth limits, something tells me a hard drive with hundreds of thousands heads would saturate this.

 

i remember reading that the rate of decay in the hardrive mageticism or whatever lead to a loss of data far faster than SSDs. I heard SSDs is in the 100+ years...i thought a couple hundred years and hardrives i thought were decades.....trying to find a source for that. you got a source?

EDIT my point was to miro gt was that i thought hardrives loose data faster than SSDs.

EDIT: granted i may be far off on this...its been a long time since i read/heard this. Also i haven't found any good sources on this

 

spinning to destiny is not the slow part, moving the head is.

 

spinning controls
primary affect is how fast data can be read/written (bandwidth)
secondary it also helps latency

heads
primary effect is how fast it can access data (latency)

 

I don't what you try to say but position the head to the right track is the bulk of the time spent in latency which is also the reason why short stroking work to compensate the laency issue.

 

just saying spinning speed also effects latency but its not the primary source for latency. 15k rpm drives have lower latency partly due to the fact that it takes less time for it to spin to a location. thats all

 

and the latency is much higher when the head has to move => data accesses massively drop. and it's that one difference where an ssd suddenly can be 100x faster and more. it's that thing that makes computers really really slow. by far the biggest impact.

 

if such device exists then it would cost lot more than a SSD would.

 

dont know about that, some of the parts have already been set up for mass production.

it would take funds for the development of the controller, but this is what also takes place with SSDs nowadays. And then back to the HDD head itself that should get smaller, thought a head itself is pretty simple unit

 

One head per side may be a simple unit, one on each track may not be. How many tracks are there ? How do you pack so many without each getting into others way ? Where would they be hang on to ?

Doesn't make sense to me, an SSD is simpler and most likely cheaper.

 

I dunno. $1000+ for a 512GB SSD is a lot of room to work with, heck even $500+ for a 256GB. Even if the tech was half the cost of an SSD with roughly half the performance it would be a better buy.

Many of us geeks here don't think twice about spending $300 for a storage drive when most others balk at spending more than $100. There's got to be some better solution that is better than the Momentus XT and a lot cheaper than conventional SSD's.

I still think a hybrid drive with user accessible 32GB or 40GB SSD in the same 2.5" unit as an HDD would be preferable. Let the user decide to use it as a "smart cache" like the XT or just use it like a separate SSD.

This would be a simple solution to the single drive laptops that make up a large majority of laptops sold, and its using technology that already exists. Just would have to come up with a controller to manage the SSD and HDD.

 

Well, space might be an issue. Platters take up a lot of room, and while the 4 GB NAND in the Momentus XT is a single chip (I think), going to 32 GB (or more) would require at least 4 chips (4x 64 Gbit ICs). It might be a little difficult to find the necessary room in a 2.5" 9.5mm standard form factor.

 

If the hard drive is a single platter then it might be possible. Don't they have 16GB NAND chips?

 

I do believe there is but I don't think I have heard of them being used in products as far as I am aware.

 

all this NAND storage on the HDD will not be necessary if this thing with the bunch of heads could be accomplished. Those will have a cache chip though.

again, it will be possible to fit more tracks on the disk. At the same time probably only one disk will be used, so we are most likely to see say a 500GB (or more) single platter HDDs with access times close to nothing and transfer speeds with times higher than those of the todays HDDs (with utilizing the use of many heads at the same time).

in other words - direct competition for the SSDs .... that offer with times the write cycles of the SSDs and long data retain !!

 

If increase couple more arms among with the heads i see the benefits in that working as RAID0 arms+heads on the same platter however being 'bulky' problem remains and will added more power consumption. And doesn't sound manufacture cost friendly.

I read alot failure of that Momentus XT, overall, just get a genuine SSD drive

 

^ the arms will not be moving anymore. The only moving thing will be the rotating data disk in the middle. So power consumption actually could drop.

 

So one straight static arm with a redesigned long head instead of multi-heads. That could reduce required hardware and using same power consumption by diverting the arm power to the head. Could work actually. But there maybe bandwidth bottleneck to the IC Board.

EDIT: Ah well, is a blasted idea anyway, if only someone able to design a single long head with ability read write disk sectors in a multi-tasking way.

 

it can't be one head though. the head is what makes the 0 and 1 by imprinting into the track so it still needs multiple heads. It could have a single head for reading i guess but definitely not for writing.

Best bet is having a single/multi-arms with several heads on it that slide up and down the arm to read but i doubt a power sliding mechanism could move fast enough.

 

Yeah I just thought if someone can design a head able to do multiple read write on a single head but that simply not possible via logical thinking.

Well then i guess having 2 heads on a single arm will do the speedup. But in the end 2 arms will do faster as one for read and one for write. Either way, that add up power consumption. Now, let's assume WD will do the prototype test of this

 

I remember reading about something like that a long time ago in some computer magazine.
Millipede memory - Wikipedia, the free encyclopedia

With the huge advance we have made when it come to lithography technology I can't imagine what kind of data density that technology could reach.

 

someone already posted that

 

the one long head is not possible, it has to be many heads on the arm, and then couple arms for improving performance.

anyway, here's what the HDD head pretty much consists of:
Disk read-and-write head - Wikipedia, the free encyclopedia

as you can see, the coil can be lithographically produced, and the rest is 4 connections - 2 for the reading part, and 2 for the writing part. There's actually two coils over that one ferrite material in the middle.

.. in other words - pretty simple. Just needs to be made smaller