Take the 2-minute tour ×
Server Fault is a question and answer site for system and network administrators. It's 100% free, no registration required.

This question bothers me for years now... If it is inappropriate here, a valid answer would point me to the perfect place.

As far as I know, striping is used in several storage situations to speed up transfer rates: multiple hdd's in one computer, chips inside of ssd's, dual/triple channel memory, and probably more, that I don't know.

Harddisks ususally have a number of platters inside. Usually 1 to 5. Sometimes they're used on both sides, which means there are up to 10 heads inside the hdd. But no matter what number of platters is inside, the transfer rates of harddisks is the same. Why?

The fact, that the heads and platters always share the same position does not count for me. The data could still be distributed between the platters, lets say one byte on each platter, of the stripe size of a regular RAID could be used as well.

What is the reason for not striping inside of harddisks, or if it is done, why is the transfer rate of a 2-platter disk not the double of a one-platter disk?

share|improve this question
    
Wouldn't a single disk then require multiple bus-connections on its controller? –  jscott Oct 30 '10 at 13:20
    
Of course, this would be completely hidden, the disk would have exactly the same type of interface as any other disk. OR: Would a SSD have multiple bus connections on its controller? –  Ingo Oct 30 '10 at 13:29

3 Answers 3

up vote 2 down vote accepted

To put simply: disks are not this precise.

What I mean, is that placement of the head over the platter (the "selected" track) and the position of the platter (the "selected" cylinder) is approximate. On the platters there are special free spaces between sectors to allow for those inaccuracies. Hitting two such places at a time would be hard.

You have to remember that we are talking about precision in the order of microns. The wind that the rotating platters create bends the read/write heads. Because some heads are experiencing vastly different wind than other (the top head and the head in the middle of the stack) trying to achieve any kind of alignment between them is nearly impossible.

share|improve this answer
    
ok, this just explains, that it is not as easy as I thought. DerfK's link to the patent shows, that there is a solution. But that must have been an issue. –  Ingo Oct 30 '10 at 13:53
    
the patent suggests using a single voice coil for each r/w head. That would make the drive complicated: not only you need to squeeze many more magnets to drive the heads, you also have to note the interference between them. It's just much harder than it looks on first sight. –  Hubert Kario Oct 30 '10 at 16:32

Possibly because whoever patented it wants too much for the license?

Another reason I can think of is that harddrives internally remap sectors when they detect that they are getting hard to read. To maintain the "stack" of sectors for striping to work, the entire stack would need to be moved, which would quickly eat up the drive's spare storage space, while wasting otherwise good sectors.

share|improve this answer
    
good idea, but I guess I would prefer to loose 10% of hdd space, and gain a factor of speed. Remember that people are paying a multiple per GB when buying 10k RPM disks. –  Ingo Oct 30 '10 at 13:43
    
Just followed the link to the patent site. I don't follow the license cost problem. The solution is there for a while (though I think the idea could have been applied decades ago), would somebody reject to earn a little money? –  Ingo Oct 30 '10 at 13:55

Another factor might be head-switching time, if there's more than one (physical) head. That's probably done by analog switches, so that that A->D conversion hardware doesn't need to be duplicated for each head.

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.