Half a croissant, on a plate, with a sign in front of it saying '50c'
h a l f b a k e r y
Assume a hemispherical cow.

idea: add, search, annotate, link, view, overview, recent, by name, random

meta: news, help, about, links, report a problem

account: browse anonymously, or get an account and write.

user:
pass:
register,


                   

Black Box RAID

A RAID without drivers
  (+1, -6)(+1, -6)
(+1, -6)
  [vote for,
against]

I propose a multi-platter hard drive in which each platter acts as a member of an internal RAID. The hard drive's own circuitry would take care of all the RAID housekeeping, leaving a single communication channel on the outside of the enclosure that is compatible with any machine of that standard (ATAPI, SATA, etc.). The throughput would only be limited to the speed of such channel. Connections for the power supply and jumpers for master/slave selection would also remain as in the current art.

One of the largest advantages is the ability to manufacture a drive with a slower spindle speed that is competitive with one that spins faster, resulting in competitive data transfer speeds with significantly less data density. Vibrations, noise, and wear would be reduced, and manufacturing tolerances could potentially be somewhat relaxed.

Special drivers could be available to perform lower-level RAID tasks (for geeks and tech support people), but no such drivers would be necessary for full RAID functionality out of the box on a standard hard drive controller.

kevinthenerd, Jul 09 2008

Cylinder-head-sector hard drive concepts at Wikipedia http://en.wikipedia...ylinder-head-sector
[phoenix, Jul 09 2008]

[link]






       Single point-of-failure
If the spindle bearing goes, you've lost the RAID. [-]

[EDIT] Thinking further, double [-], unless you propose multiple heads and head actuators, otherwise a bad sector or track on a single platter would require a seek away from the cylinder that all the other heads are happily reading/writing from/to.

Besides which, of course, this idea was baked by Fujitsu in the 1980s. And probably invented by a Frenchman.
coprocephalous, Jul 09 2008
  

       RAID stands for Redundant Array of Inexpensive Disks.   

       Now your idea:   

       Redundant - nope   

       Array - uh uh   

       Inexpensive - no sireee   

       Disks - disk yes? disks no.   

       All of the advantages of RAID are missing from this idea.
Giblet, Jul 09 2008
  

       sadly it wouldn't work, I've suggested this before to people and apparently the temperature variations within the disk drive means all the platter heads wouldn't line up accurately enough.   

       I've got a 5.25" 4200rpm HD (last model produced in a 5.25" form factor) and while I do like it and wouldn't trade it, it's not as quiet as you'd imagine: while the motor noise is lower, the wind noise is higher.
FlyingToaster, Jul 09 2008
  

       No, I think this will work. [coprocephalous] is right there is still a single point of failure, and bad blocks would block use of paired blocks but I think it could double or triple sequential read/write times and that could be so useful people would deal with the downside.   

       [FT]'s note about misaligned heads confuses me. Are you saying that the center platter in a three platter drive might expand more in the heat and that has to be accounted for in head positioning? So you couldn't stream part of a bit to the three platters simultaneously? If that's true it's a dealbreaker for 2+ platter disks. But even then I think it would work for 2 platter disks.
MisterQED, Jul 09 2008
  

       Most hard drives do have multiple platters and multiple heads which read and write simultaneously across the cylinder. The cylinder is the track position on a given platter projected through the all the platters.   

       Contrary to your assertion, hard drive data transfer speeds are limited largely by the mechanical aspects of the drive. That is, you can only read or write data as fast as it passes under the read/write heads. A slower spindle speed, without some sort of predictive caching, will always result in slower transfer speeds.
phoenix, Jul 09 2008
  

       I'm not a HW expert at that sort of thing, but I just looked up a recent disk-drive model available in 5 different capacities... they each have the exact same disk>cache transfer rate even though they have different number of heads/platters.   

       I imagine (but don't really know) that even though you would eventually reach an equilibrium where each platter was at its own constant temperature, it might take awhile to get there; if you "raid'd" the platters you might not be able to use the disk until all the platters had settled in.   

       However I would be more than happy to accept a conspiracy theory of some sort, ie: interfaces can't handle the speed, saving it for a marketing rainy day, etc.
FlyingToaster, Jul 10 2008
  

       What Giblet said
Voice, Jul 10 2008
  

       ie: the word you're looking for is "striped"
FlyingToaster, Jul 10 2008
  
      
[annotate]
  


 

back: main index

business  computer  culture  fashion  food  halfbakery  home  other  product  public  science  sport  vehicle