Overclock.net › Forums › Components › Hard Drives & Storage › idea: one head per track on a HDD?
New Posts  All Forums:Forum Nav:

idea: one head per track on a HDD? - Page 2

post #11 of 20
Quote:
Originally Posted by LemonSlice View Post
It doesn't work like that. If you lose one head, the entire drive stops functioning. You don't just lose the data on that head. The only way to get the data back is expensive data recovery methods and those can recover data from bad heads most of the time if the disk remains intact.

So the more heads you have, the more chance of one out of the bunch failing. When that one head fails it will take out the whole drive. At least I've never heard of a hard drive that can run off a dead head. Also the heads are indeed very small, but they are mounted on a moderately large piece that is designed to actually lift the heads off the disk just a microscopic amount with the air flow inside the disk. This piece is roughly 1x1mm and there is no way to fit enough of these across the disk to read all the tracks simultaneously.
are you bad at reading? I said ASSUMING IT DOESN'T TAKE OUT THE WHOLE DRIVE. basically saying that they have resigned the circuitry anyway for a drive like that, they could have easily made is so that a failed head doesn't make the drive useless. just like on a triple axle trailer, if one tire fails, you can still drive it.

Maybe you should think outside the box for once in your life, this thread is just about brainstorming and talking about possible idea, so pull the stick out of your ass and calm down.
post #12 of 20
Quote:
Originally Posted by thrasherht View Post
are you bad at reading? I said ASSUMING IT DOESN'T TAKE OUT THE WHOLE DRIVE. basically saying that they have resigned the circuitry anyway for a drive like that, they could have easily made is so that a failed head doesn't make the drive useless. just like on a triple axle trailer, if one tire fails, you can still drive it.

Maybe you should think outside the box for once in your life, this thread is just about brainstorming and talking about possible idea, so pull the stick out of your ass and calm down.
Calm down? I haven't gotten angry about anything anyone has said, I'm simply providing facts that I know. Instead of getting angry about me stating facts, why can't you accept my ideas about why it won't work as valid comments too? Judging by your first comment, it is either you who can't proofread or just proved the idea you're supporting is not effective: "I can imagine if you could implament this correctly it would greatly increase the seek time of drives. the only problem i see, is the mast of circuitry needed to run all of the heads."

Lets say it wouldn't take out the whole drive though; since there are so many heads, if one fails, you will still lose data. Losing less data is not a good reason to increase the change of losing said data. Also hard drive motors actually have very little torque, so if there are a lot of heads resting on top of the disk, it probably wouldn't spin up, at least not the current motors. In a hypothetical drive with 65,536 cylinders, 16 sectors per track, and totaling 127.5GB, lets assume there are 3 platters, though probably not nearly that many at today's drive density standards. That would be 6 faces of platters. Cylinders are collections of tracks that lie on the same plane. So since there are 6 faces, there would be a need for about 10,923 heads per face of a platter.
Edited by LemonSlice - 12/22/10 at 4:06am
Tiny Fun
(8 items)
 
4P Folding Rig 1
(16 items)
 
4P Folding Rig 2
(11 items)
 
CPUMotherboardGraphicsRAM
i7-4770K @ 4.0GHz MSI Z87I EVGA GTX 660 Ti Corsair Vengeance 16GB 
Hard DriveCoolingOSCase
Samsung 840 Pro 256GB Swiftech Apogee Drive II Windows 8 Enterprise x64 Cooler Master Elite 120 Advanced 
CPUCPUCPUCPU
AMD Opteron 6274 ES 2.2GHz 16-Core AMD Opteron 6274 ES 2.2GHz 16-Core AMD Opteron 6274 ES 2.2GHz 16-Core AMD Opteron 6274 ES 2.2GHz 16-Core 
MotherboardGraphicsRAMHard Drive
SuperMicro H8QGi+-F Matrox G200eW 16x2 (32GB) Crucial Ballistix DDR3 1600MHz Crucial C300 64GB 
OSPowerCase
Linux Mint 15 Cinnamon 64-bit Enermax Galaxy Evo 1250W Spotswood Tech Tray 
CPUCPUCPUCPU
AMD Opteron 6176 SE 12-Core 2.3GHz AMD Opteron 6176 SE 12-Core 2.3GHz AMD Opteron 6176 SE 12-Core 2.3GHz AMD Opteron 6176 SE 12-Core 2.3GHz 
MotherboardGraphicsRAMHard Drive
Supermicro H8QGL-IF+ Matrox G200eW Kingston Server 16x1GB DDR3 Corsair Nova 32GB 
OSPower
Linux Mint 15 Cinnamon 64-bit Cooler Master Silent Pro 1000W 
  hide details  
Reply
Tiny Fun
(8 items)
 
4P Folding Rig 1
(16 items)
 
4P Folding Rig 2
(11 items)
 
CPUMotherboardGraphicsRAM
i7-4770K @ 4.0GHz MSI Z87I EVGA GTX 660 Ti Corsair Vengeance 16GB 
Hard DriveCoolingOSCase
Samsung 840 Pro 256GB Swiftech Apogee Drive II Windows 8 Enterprise x64 Cooler Master Elite 120 Advanced 
CPUCPUCPUCPU
AMD Opteron 6274 ES 2.2GHz 16-Core AMD Opteron 6274 ES 2.2GHz 16-Core AMD Opteron 6274 ES 2.2GHz 16-Core AMD Opteron 6274 ES 2.2GHz 16-Core 
MotherboardGraphicsRAMHard Drive
SuperMicro H8QGi+-F Matrox G200eW 16x2 (32GB) Crucial Ballistix DDR3 1600MHz Crucial C300 64GB 
OSPowerCase
Linux Mint 15 Cinnamon 64-bit Enermax Galaxy Evo 1250W Spotswood Tech Tray 
CPUCPUCPUCPU
AMD Opteron 6176 SE 12-Core 2.3GHz AMD Opteron 6176 SE 12-Core 2.3GHz AMD Opteron 6176 SE 12-Core 2.3GHz AMD Opteron 6176 SE 12-Core 2.3GHz 
MotherboardGraphicsRAMHard Drive
Supermicro H8QGL-IF+ Matrox G200eW Kingston Server 16x1GB DDR3 Corsair Nova 32GB 
OSPower
Linux Mint 15 Cinnamon 64-bit Cooler Master Silent Pro 1000W 
  hide details  
Reply
post #13 of 20
Thread Starter 
if we replace the 10,923 1*1mm heads by a huge one that covers the 10923 tracks and can read every one of these tracks simultaneaously without moving, it would actually work...

and the fact that there is no moving parts (except the drive spinning) could improve it's reliability...

also, we managed to realize sooo many thing we first thought impossible to do... you have to agree with me that this COULD work one day... And it would definitely beat SSDs! similar/faster speeds and much more storage
    
CPUMotherboardGraphicsRAM
I7 920D0 @4,2Ghz HTon1,29V #3844A937 Gigabyte EX58-UD4P GTX 570 Patriot Extreme Viper 6GB DDR3-1600 8-8-8-24 
Hard DriveOptical DriveOSMonitor
2xSeagate 7200.12 500Gb RAID0 LG GH-22 Windows 7 Ultimate x64 Syncmaster P2370 -- 23" 
PowerCase
Corsair 750W XClio A380BK 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
I7 920D0 @4,2Ghz HTon1,29V #3844A937 Gigabyte EX58-UD4P GTX 570 Patriot Extreme Viper 6GB DDR3-1600 8-8-8-24 
Hard DriveOptical DriveOSMonitor
2xSeagate 7200.12 500Gb RAID0 LG GH-22 Windows 7 Ultimate x64 Syncmaster P2370 -- 23" 
PowerCase
Corsair 750W XClio A380BK 
  hide details  
Reply
post #14 of 20
Thread Starter 


the best picture i was able to find... but i don't know the age of this head...
    
CPUMotherboardGraphicsRAM
I7 920D0 @4,2Ghz HTon1,29V #3844A937 Gigabyte EX58-UD4P GTX 570 Patriot Extreme Viper 6GB DDR3-1600 8-8-8-24 
Hard DriveOptical DriveOSMonitor
2xSeagate 7200.12 500Gb RAID0 LG GH-22 Windows 7 Ultimate x64 Syncmaster P2370 -- 23" 
PowerCase
Corsair 750W XClio A380BK 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
I7 920D0 @4,2Ghz HTon1,29V #3844A937 Gigabyte EX58-UD4P GTX 570 Patriot Extreme Viper 6GB DDR3-1600 8-8-8-24 
Hard DriveOptical DriveOSMonitor
2xSeagate 7200.12 500Gb RAID0 LG GH-22 Windows 7 Ultimate x64 Syncmaster P2370 -- 23" 
PowerCase
Corsair 750W XClio A380BK 
  hide details  
Reply
post #15 of 20
Quote:
Originally Posted by Lome_Wolf236 View Post
if we replace the 10,923 1*1mm heads by a huge one that covers the 10923 tracks and can read every one of these tracks simultaneaously without moving, it would actually work...

and the fact that there is no moving parts (except the drive spinning) could improve it's reliability...

also, we managed to realize sooo many thing we first thought impossible to do... you have to agree with me that this COULD work one day... And it would definitely beat SSDs! similar/faster speeds and much more storage
Still couldn't beat a SSD. You are attempting to use moving physical parts to beat the speed of electron. i.e. The fastest mechancial adding machine cannot beat an simple electronic calculator.


However, the concept of the Hard Rectangular Drive is similar to what you propose:
http://arstechnica.com/hardware/news...-hard-disk.ars

Once again...
(13 items)
 
  
CPUMotherboardGraphicsRAM
i7 920 [4.28GHz, HT] Asus P6T + Broadcom NetXtreme II VisionTek HD5850 [900/1200] + Galaxy GT240 2x4GB G.Skill Ripjaw X [1632 MHz] 
Hard DriveOSMonitorKeyboard
Intel X25-M 160GB + 3xRAID0 500GB 7200.12 Window 7 Pro 64 Acer H243H + Samsung 226BW XARMOR-U9BL  
PowerCaseMouseMouse Pad
Antec Truepower New 750W Li Lian PC-V2100 [10x120mm fans] Logitech G9 X-Trac Pro 
  hide details  
Reply
Once again...
(13 items)
 
  
CPUMotherboardGraphicsRAM
i7 920 [4.28GHz, HT] Asus P6T + Broadcom NetXtreme II VisionTek HD5850 [900/1200] + Galaxy GT240 2x4GB G.Skill Ripjaw X [1632 MHz] 
Hard DriveOSMonitorKeyboard
Intel X25-M 160GB + 3xRAID0 500GB 7200.12 Window 7 Pro 64 Acer H243H + Samsung 226BW XARMOR-U9BL  
PowerCaseMouseMouse Pad
Antec Truepower New 750W Li Lian PC-V2100 [10x120mm fans] Logitech G9 X-Trac Pro 
  hide details  
Reply
post #16 of 20
Thread Starter 
it's the same thing, but instead of going back and forth 500 times/second ( i think) it would still be a hard drive, spinning at 5400 or 7200 rpms, and because the heads wouldn't move, the latency would be greatly reduced!


SSDs will always beat hard drives in latency, but this idea would allow hard drives to reach SSDs bandwith, and they could be faster using better controllers

7200rpm == 120 rotations per second == 8.33ms/rotation
since there is one head/track, and if my logic is right, it could read the whole drive in 9ms, IN THEORY, but still do want?
it will only be limited by the controllers and SATA bandwidth, and increasing the rpm will increase it's theorical limit...

( i do know theorical limits are never reached, but we are far away from seeing a 2TB SSD for 100-200$ with that potential)
    
CPUMotherboardGraphicsRAM
I7 920D0 @4,2Ghz HTon1,29V #3844A937 Gigabyte EX58-UD4P GTX 570 Patriot Extreme Viper 6GB DDR3-1600 8-8-8-24 
Hard DriveOptical DriveOSMonitor
2xSeagate 7200.12 500Gb RAID0 LG GH-22 Windows 7 Ultimate x64 Syncmaster P2370 -- 23" 
PowerCase
Corsair 750W XClio A380BK 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
I7 920D0 @4,2Ghz HTon1,29V #3844A937 Gigabyte EX58-UD4P GTX 570 Patriot Extreme Viper 6GB DDR3-1600 8-8-8-24 
Hard DriveOptical DriveOSMonitor
2xSeagate 7200.12 500Gb RAID0 LG GH-22 Windows 7 Ultimate x64 Syncmaster P2370 -- 23" 
PowerCase
Corsair 750W XClio A380BK 
  hide details  
Reply
post #17 of 20
How on earth do you think you will manage to read the whole drive at once? Triple channel DDR3 maxes out at less than 50GB/s, yet somehow your drive is going to manage over 20TB/s? I think you're getting a little carried away (especially on the pricing - do you really think it's going to be cheap to produce such a drive, plus a controller that can exceed i7 bandwidth, let alone SSD controller bandwidth?)
post #18 of 20
Thread Starter 
read my post again (properly), i said that was THEORY and only THEORY
i'm just saying that it could easily max out SATA standard (3Gb/s or even 6 Gb/s)

20Tb/s? who said it would deliver that bandwidth TODAY? but it's definitely possible to reach that performance.

the most expensive part of current HHDs are the platters, i guess. now add a ton of heads (which will bring down their price due to mass production) and a modified SSD controller (optimized for this HHD), and i'd say it would not cost over 300$ if produced in decent quantities. But it would have a MUCH better price/Gb than SSDs with similar performance


EDIT: where is the 20Tb/s coming from anyway? it's more like 120*(size of the drive) for a 7200rpm drive
Edited by Lome_Wolf236 - 12/23/10 at 7:21am
    
CPUMotherboardGraphicsRAM
I7 920D0 @4,2Ghz HTon1,29V #3844A937 Gigabyte EX58-UD4P GTX 570 Patriot Extreme Viper 6GB DDR3-1600 8-8-8-24 
Hard DriveOptical DriveOSMonitor
2xSeagate 7200.12 500Gb RAID0 LG GH-22 Windows 7 Ultimate x64 Syncmaster P2370 -- 23" 
PowerCase
Corsair 750W XClio A380BK 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
I7 920D0 @4,2Ghz HTon1,29V #3844A937 Gigabyte EX58-UD4P GTX 570 Patriot Extreme Viper 6GB DDR3-1600 8-8-8-24 
Hard DriveOptical DriveOSMonitor
2xSeagate 7200.12 500Gb RAID0 LG GH-22 Windows 7 Ultimate x64 Syncmaster P2370 -- 23" 
PowerCase
Corsair 750W XClio A380BK 
  hide details  
Reply
post #19 of 20
Possible, yes.
Practical at a sensible price - no.

Some of the big old mainframe hard drives had multiple heads per platter, usually 2-5 I think. A dual head solution with one head on the inner half of the drive, and one on the outer, wouldn't be insanely expensive and would help seek times, they could even be able to seek the whole platter separately. It would drive the price up a bit, but if you really want ultra fast, then solid state is really the only way to go.
Ryzen 5 1600
(12 items)
 
  
CPUMotherboardGraphicsRAM
AMD Ryzen 5 1600 MSI B350 Gaming Plus Gigabyte GeForce GTX 1060 WINDFORCE2 OC 16GB (2x8GB) Corsair DDR4 Vengeance LED, PC4-24... 
Hard DriveHard DriveCoolingOS
256GB Samsung PM961 Polaris M.2 NVMe  1TB Toshiba DT01ACA100 3.5" HDD, SATA III  Cooler Master Hyper 212 Evo Windows 10 64 
KeyboardPowerCaseMouse
Unicomp Model M 650W EVGA SuperNOVA G1, 80PLUS Gold, Full Modular Kolink Luminosity Cooler Master Reaper Aluminium 
  hide details  
Reply
Ryzen 5 1600
(12 items)
 
  
CPUMotherboardGraphicsRAM
AMD Ryzen 5 1600 MSI B350 Gaming Plus Gigabyte GeForce GTX 1060 WINDFORCE2 OC 16GB (2x8GB) Corsair DDR4 Vengeance LED, PC4-24... 
Hard DriveHard DriveCoolingOS
256GB Samsung PM961 Polaris M.2 NVMe  1TB Toshiba DT01ACA100 3.5" HDD, SATA III  Cooler Master Hyper 212 Evo Windows 10 64 
KeyboardPowerCaseMouse
Unicomp Model M 650W EVGA SuperNOVA G1, 80PLUS Gold, Full Modular Kolink Luminosity Cooler Master Reaper Aluminium 
  hide details  
Reply
post #20 of 20
Quote:
Originally Posted by Lome_Wolf236 View Post
read my post again (properly), i said that was THEORY and only THEORY
i'm just saying that it could easily max out SATA standard (3Gb/s or even 6 Gb/s)
But it isn't THEORY, since it is built on no evidence nor does it offer any explanation. What you have is an IDEA.

A few more points - a head is MUCH bigger than a single track, but the magnetic field that does the reads/writes is focused on the platter. There is no way to (currently) do 1 head per track without reducing areal density. If you put a single head per track, you get very limited data storage per head, with poor areal density per platter. So your storage capacity drops, but your costs go up.

Now think about channels. Hard drive controllers are single channel. MLC controllers are 4-10 channel. In order to read from all your heads concurrently, you are asking for a multi-thousand channel controller. How much do you think that's gonna cost, and (just as importantly) how do you think you're gonna wire it all up? Even the Hard Rectangular Drive limits itself to 64 heads, and that is in pricing-undisclosed equipments targetted at the enterprise market. Still think you're in SSD-beating budget territory?

Now on to latency. Unless you have a buffer for all that data you read in a single pass (and to do so would be stupid, since you would then have a buffer for the full disk, which would be even more cost prohibitive), you will have to rely on the stored bits passing under the heads. For 7200rpm disks, that means an average latency of just over 4ms per request. So your seek time is now zero, but your access time is still 4ms+. So even though you get (the potential) for massive transfer rates, you still have terrible latency, which is easily bested by SSD technology.


Quote:
Originally Posted by Lome_Wolf236 View Post
20Tb/s? who said it would deliver that bandwidth TODAY? but it's definitely possible to reach that performance.
Sure it will be. But it will be (likely) easier and cheaper to do so with other technologies - most likely solid state chips. And I can't see it happening in the near future, since the bandwidths are so far in excess of what is currently available.

Quote:
Originally Posted by Lome_Wolf236 View Post
he most expensive part of current HHDs are the platters, i guess. now add a ton of heads (which will bring down their price due to mass production) and a modified SSD controller (optimized for this HHD), and i'd say it would not cost over 300$ if produced in decent quantities. But it would have a MUCH better price/Gb than SSDs with similar performance
Hard drives currently sell millions of units a year, and as each unit has 1-10 heads already, do you really think mass production hasn't brought down the cost per head about as far as it will go?

Quote:
Originally Posted by Lome_Wolf236 View Post
where is the 20Tb/s coming from anyway? it's more like 120*(size of the drive) for a 7200rpm drive
You're right - I was typing in a hurry. The bandwidth you are looking for is in excess of 240TB/s as you mentioned - you're previous post compared your new option to a 2TB SSD, and required reading the entire capacity in a single platter rotation.

Now I don't like tearing people's ideas apart, as it is only through ideas such as these that we get innovation. But your being kinda aggresive about the whole thing. When people bring up valid obsticles to your design you would be better placed if you think of possible ways around the problems raised (such as some form of parity protection to avoid the head failure issue resulting in data loss as discussed before) rather than jumping on people and accusing them of either not reading or understanding properly.
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Hard Drives & Storage
Overclock.net › Forums › Components › Hard Drives & Storage › idea: one head per track on a HDD?