It's an ideea I've had for some time now.
Is it possible to put two smaller SATA hard-drives,link them up in RAID 0,somehow jam them into a 3.5inch shell that would look just like a normal hard-drive only it would be far faster?.Also,there's the point of somehow making the motherboard recognize the drive.
Nevertheless is it possible?Performance related,it's not such a bad ideea.

it would cost a lot... and probably wouldn't have stellar performance... because with most smaller HDD they don't perform as well as some of the larger ones.... and say a velociraptor wouldn't work because then your getting really high prices

Well Paul, what you mentioned are dual actuator drives, and Conner Peripherals (now defunct) produced them almost twenty years ago. The drives were called the "chinook" line of hard disks.

Think about that, western digital is saying that it would "cost too much" with today's miraculously cheap drive prices, in the days of $500 dollar video cards, and 900 dollar extreme edition processors.

They are also saying it's complex and would raise failure rates...so what ? With the weak link in performance being the hard drive, and people buying 300 dollar raptors, does this make sense to you ? Just like any hi performance bleeding edge part, the likelihood of failure is greater than your mid range products.

Consider also that seek times would be halved on average as well.

I think it is pretty clear that drive manufacturers are too lazy and / or lack the will to do it. We certainly have the tech at reasonable prices, however it seems that they are of the attitude, "you'll buy only what we feel like making and you'll buy our bs reasons and like it."

Tom

One idea I came up with when thinking about this whole arm idea... why not replace the moving arm with something really cool.

Imagine if the head covered the entire length from the inside cylinder to the outside one and instead of moving a head back and forth you digitally select the area of the drive you wish to read. I think this would be a massive performance increase and also increase reliability (then if you could multiplex the signals coming off the head you could read from multiple area's at the same time)

Anyone who would like to use my idea may do so and if you feel generous you could throw me a few bucks lol

Well, cionsidering the advancement in manufacturing and electronics, i can't really buy the difficulty argument when they did it ears ago:

http://www.storagereview.com/guide [...] tiple.html

http://en.wikipedia.org/wiki/Conne [...] ator_drive

I think we need Tom's and Annand to really put them on the hotseat on this one and cough up specifics...or the REAL reason(s).

Have a good weekend man.

yah, you're thinking like me, a square cd with a mesh of leds/ lasers to read it. I think the main problem would be manufacturing a read head of awesome density and quality control. Sure beat what we have now.

Hey Paul, just fyi...a regular vanilla cable line at 8mhz freq (2 up 8 down) has a down stream bandwith of about 58 megabytes per sec. Last i checked this was back in the docsis 1.xx days...so the hd really is the weak link.

That was the 2305 drum http://www-03.ibm.com/ibm/history/ [...] _2305.html
A very expensive device.

I think we might see some sort of a hybrid flash/hard drive device.

Rather than use dual read/write actuators (which would use a lot of space inside the drive and increase moving parts, which increases failure probabilities), it would be interesting to investigate a hard drive design that keeps the single read/write actuator, but wires half of the heads to one read/write processor, and half to a second read/write processor.

This way, on a 4-platter drive (8 heads), wire the 4 heads on the tops of the platters to one read/write processor, and the 4 heads on the bottoms of the platters to a second read/write processor. Allow the processors to run in parallel, reading information from both sides of a platter simultaneously, and interleave the read data for transmission to the computer (like RAID 0 would do).

I'm not sure why this hasn't been done ... although I suspect there may be some significant processing overhead as the drive decides how to interleave data on reads and split up data on writes, as well as properly synchronize the two controllers.

The reason it hasn't been done is that they aren't precisely aligned enough - the tracks are so narrow that there is no way to keep both the top and bottom head on a track at once. One is following the track, and the other is slightly off.

Ah, interesting. That would mean you would have to have independent tracking of each group of heads.

While this would seem to imply that you need two independent actuators, I think there's a way around that.

Say you're wanting to read from both the top and bottom of the platter with 2 heads simultaneously. You align and begin reading using the top head. Now the bottom head is not aligned, but it's very close. You don't really need to independently be able to move the bottom head over the entire platter, you just need to be able to slightly shift its alignment relative to the top head to get both heads aligned at once.

Currently, for a 2-platter drive, all 4 heads are attached together on one head assembly, a rigid structure. Split this into two assemblies, one with the two top heads, one with the two bottom heads, and a mechanical joint between them that can allow the two structures to move relative to each other by a very small amount, say 5 microns. Now, the entire structure moves together using the voice coil, but the top/bottom head assemblies can move relative to each other using a very small piezoelectric actuator. Thus, your top head processor now aligns the top heads using servo control via the voice coil like normal, and your bottom head processors align the bottom head using the piezoelectric actuator.

I think this could easily fit inside the currently designed 3.5" enclosure and still use only one voice coil mechanism and actuator.

The only thing with this design is that you won't double your average throughput. It will now take longer to align the bottom heads because the top head has to settle before you can fine-align the bottom one. It ends up increasing access time to the logical blocks on the bottom of the platters. For large files, the throughput will be close to double, but for many small, random reads, the throughput will go down compared to a standard drive.

Damn...took me about an hour to read and understand this,but I get it.Noob,so be gentle .In other words,drive manufacturing is like politics : we have the technology,the money,the time...just not the will.
A little off-topic,I've read a few years back that some company (probably WD) produced a holografic 1.8TB drive.Any NFO on that?Also,I've heard recently of a decent sized pure flash drive which used the 40-pin ATA connection method but had dramatically decreased seek times.
Thanks for the replies!
Love&Peace!
P.S. : This whole thing that WD said that puting two armatures in one drive would increase the probability of a failure (even though it's 0.3% per year;I think I read this on the Seagate homepage)in the drive reminds me of a HDD that I have in storage,somewhere.It's a Seagate,it's big (bigger than 5.5').Manufactured on May 20th 1984-1985 ( can't remember).And this is what's written on it : Warning!Fragile!Do not subject to forces greater than 350Gs or waranty will be void

Putting in two sets of actuators with heads they can position only addresses half of the seek time. When a sector is being sought, the first phase is to position the heads over the right tracks (collectively, one cyclinder), turn "on" the right head, check that it is aligned with the track and start reading what flys past it. The next phase is to watch the data coming from the head and wait until the right sector comes by to yield its data. This last phase may take from almost none to all of one revolution of the platters - on average it will take one half revolution. That is why high-rpm disk like Velociraptors have faster access - they shorten this particular phase of the seek process. Modern drives with large buffers will actually read into the buffer MANY sectors in sequence after the first one requested, in hopes that the next one needed will be the next one on the disk and it will already be in the buffer when requested. Periodic defragmentation of the drive (to ensure file pieces are in sequence on the disk) helps this a great deal. I grant you that putting two actuator / head assemblies on opposite side of the platters would even shorten the platter-turning phase (because you only have to wait for up to half of a revolution) PROVIDED THAT the software watches BOTH heads on that track and grabs the first instance of finding the sector being sought.

Re SomeJoe7777's idea of parallel processing of pairs of heads (top and bottom) on a single actuator assembly (a little like RAID0 interleaving), MAYBE the head alignment issue is not as bad as we think. Simultaneous perfect alignment of two separate heads to existing tracks is very difficult, but just suppose that, once the two heads are installed, their relative alignment NEVER changes. Now the "exisiting tracks" simply do NOT exist as the HDD is manufactured. They are CREATED as a first step in low-level formatting at the factory, and that would be done with those same heads! So as long as the two heads remain in fixed relative alignment, the tracks they created also will remain aligned. Anyone know whether heads do this now? Maybe nobody has ever asked, because as HDD's work now the question is useless.