h a l f b a k e r yWarm and Fussy
add, search, annotate, link, view, overview, recent, by name, random
news, help, about, links, report a problem
browse anonymously,
or get an account
and write.
register,
|
|
|
Please log in.
Before you can vote, you need to register.
Please log in or create an account.
|
The fastest hard drives have a maximum read/write performance of close to 90MB/sec. These drives use the latest technology; high rotation speed, perpendicular recording to maximize data density, etc. Despite these improvements, the mechanical nature of the device makes it the slowest component in a computer
system.
Some limitations are due to form factor, such as disks being restrained to 5” diameter and a standardized height that limits the number of platters. Others are due to the actual design of the drives. In fact all hard drives have a common shortcoming – they only have one actuator arm on which the read/write heads are carried. This is a major limitation. By increasing the number of actuators, one can double, triple or even quadruple the performance of the drive.
A drive with two actuators (Dual Heads) would have twice the read / write speed of a drive with a single head. Such a drive is capable of multitasking, as the heads can work independently of each other, giving two applications maximum performance simultaneously. Even better than that, one set of heads can be reading while the other is writing. Adding more actuators, up to a total of four per drive (Quad Heads), would serve to further increase both multitasking and throughput.
With Quad Heads, data transfer speed should increase by a factor of eight. New performance tests would be required to accurately gauge disk performance, as current benchmarks only measure read and write as independent rather than simultaneous operations. With a simultaneous data transfer rate of 360 MB/sec. for reads AND writes, Quad Head technology would render SATA 3.0, (approx. 300MB/sec.) and the proposed SATA 6.0 (600MB/sec.) obsolete overnight.
Like this?
Multiheaded_20CD-ROM
Posted back in 2001. [st3f, Feb 20 2007]
Disk Failure Analysis
http://216.239.37.1...s/disk_failures.pdf
Failure Trends in a Large Disk Drive Population [nuclear hobo, Feb 20 2007]
Dual Hard Drive
http://i37.tinypic.com/dww5mf.png
This is an idea I put together one day, although it has only two read arms instead of four, this has the advantage of being the same width as a standard drive. [toastmonster, Sep 28 2008]
[link]
|
| |
Bun, an unqualified, *I know only the vaguest faint traces of a clue about this* bun, but a bun nonetheless. |
|
| |
I never understood why drives have only one read head. It is my understanding that the drive spins at a constant speed, and the head articulates linearly out/in. This begs the question of why not to incorporate more read heads, and for that matter, why stop at four? is there some architectural issue that prevents six or eight? The microdrives featured on I-pods, etc must have suitably miniaturised components. |
|
| |
One question, from a process-efficiency freak like myself: where is the bottleneck? Is it bulk read/write, search times, peak data rates, etc? Can faster data rate buffers be used to attenuate the bottleneck; such as onboard RAM or flash memory <if flash memory is faster; can't remember>?? Ie larger buffers than are currently available - it is clear that for this issue, cost is not the most important factor. |
|
| |
Oh yeah, not sure why the transfer speed increases by a factor of eight not four. is that a firm figure? |
|
| |
If its as simple as it sounds, I'm surprised its not already in use. |
|
| |
I like the idea! It would take some fancy drivers to allow parallel reads and writes on the same drive...but that's not a problem. |
|
| |
Only problem I see is double chance of drive failure, since two heads to fail, and expense. But expense isn't bad considering the performance gains you would get. |
|
| |
I think the current bottleneck is the transfer speed from the hard disk into the computer system, all the way to the CPU. There's just limitation in the bandwidth available for the data to travel 'all that way', as it were. The idea behind RAID 0 is that you now have more pipelines to access the same data, and so therefore it's faster. |
|
| |
Having multiple heads, each possibly dedicated to certain ranges of radii on the platter would likely reduce seek time. I think current common seek times are on the order of a few milliseconds, but that can be long in terms of computer time. |
|
| |
I think nomel is right though, the chance of drive failure (often caused by head crashes) rises dramatically, and the performance gain may not be worth the risk. |
|
| |
//not sure why the transfer speed increases by a factor of eight not four// |
|
| |
Four would be a linear measure, a factor had to be added for true simultaneous read/write. Also, multiple heads would allow operation similar to RAID 0, providing an additional speed increase. |
|
| |
//the current bottleneck is the transfer speed from the hard disk into the computer system// |
|
| |
The bottleneck is drive performance. SATA 3.0 provides 300MS/sec. bandwidth. The fastest drives are only good for about 90MB/sec. |
|
| |
//double chance of drive failure// |
|
| |
The MTBF for Seagate Savvio drives is 1,400,000 hrs. Even if the addtional complexity of four heads reduced reliability by 75%, MTBF would stil be on the order of 350,000 hours. |
|
| |
I think that the main reason this is not
done is that the head (or rather the arm
that moves it around) is the most fragile
and failure-prone part of the harddrive,
due to its mechanical nature. Having n
heads thus essentially boils down to
dividing the drive's life expectancy by n. |
|
| |
What about one of the other heads taking over from a failed one? This would actually increase MTBF, at least as far as giving you the chance to retrieve the data with the remaining good head when one or more fail. |
|
| |
What if one head is reading something while another head is writing to it? Or writes to something another head is in the middle of reading? Wouldn't preventing these occurrences slow stuff down, or at least use processing power? |
|
| |
What's not baked about multiple actuators in hard disks?
Fujistsu produced these years ago, and were in use in Quantel Crystal devices in the 1980s, and probably most early video hard disk storage devices. |
|
| |