single hard drive raid

would you be interested in this?


  • Total voters
    217

malcolm

Newcomer
Would it be possible for the seperate platters of a hard drive to work like multiple hard drives in raid? (the one im interested in is raid0 mode)
I want raid only for the extra speed, not for the extra space of the multiple drives, i dont need that much, i realy dont know what im gonna do with 320GB.
I guess it isnt possible, otherwise they would exist.
Or is it just to complicated to do for now? Are they working on it?
A raid hard drive with 4 platters could have sustained 150MB/s serial ata data rate.
I think a lot of people would be interested in this, im asking the poll to know for sure :D
If what im saying is just stupid, just ignore the poll.
 
Don't drives already do this? That's why drives with lots of platters I believe there was this 80gig with 4 20gig platters that needed ATA 66, since it would max out ATA33 to reach it's full potential. I think it was from IBM.

Now, if you're talking about basically multiple indepent heads reading around on multiple platters, well that's overly mechanically complex which means more things can break down. That's no good, besides the cost is prohibative. A 80 gig drive is 160 CDN and a 40 gig drive is 130 CDN, both 7200 RPM, what's really the point? Getting an IDE raid setup right now is pretty damn cheap, considering. Of course, when the heck do you need that much throughput? It's all about burst performance and that's where you'd probably apperciate those 8 meg caches.
 
So youre saying the same series harddrive with more platters is already faster?
How much faster is a 3-platter drive compared to 2-platter?
On tomshardware they were talking about reliability problems with multiple drives in raid0, can anyone give me some numbers on this?
Like whats the chance of 2-3-4 drive raid crashing in 1000days 10 hours a day? ( something like that )
I realy have absolutely no idea :LOL:
 
You should be able to find this information out about you favourite HDD, at which point linearly add the probability to arrive at the higher probability of drive failure.
 
Go to storagereview.com for information about HD speeds, etc.

Afaik, you can't RAID a single HD. It requires 2+. The probability for drive failure or rather data loss for a 0 type of setup is linear as Saem points out because in that type of setup, your data is split across multiple drives so if even one dies, your data is lost even though the other drive is fine.
 
Technically they could do it, practically the market probable would not bear the extra costs.
 
Hi,

The reason that you don't see many drives that do internal stripping stems from the fact that in most modern HDs, all of the heads are mounted on a single actuator. What this basically means is that any time a seek is required, the entire actuator arm moves causing the heads to move as well. If the platters were perfectly aligned, it wouldn't be a problem to do reads/writes of the same track on each of the platters in parallel, but unfortunately this really isn't possible. As platters spin, they actually change shape based on how fast they are spinning. In addition, platters at the top of the drive tend to have a different temperature than platters at the bottom of the drive. As a result, it would be fairly disasterous to try and read data from each head on the same actuator.

A way to get around this is to use multiple actuators, and have multiple heads per platter. This is what the seagate ST12450W did about 5-6 years ago. It has 2 heads per surface, with 4 heads per platter, using 2 actuators to stripe the data. It was only a 7200 rpm drive with fairly small platter density, but was only beaten 3-4 years later with the advent of the 10K rpm cheetah drives. The problem was that it's not cheap to add new actuators, heads, logic, etc, and the drive ended up not being very economcial so seagate eventually canceled it.

Now having said all of this, having more platters even without reading them in parallel can increase drive performance, but it's really a tradeoff. On one hand, having a lot of platters means that (at the same platter density) you don't need to move the actuator as often for track-to-track seeks. Instead, you are doing head switches which are a fair deal cheaper. This results in faster seek times. At the same time though, having a lot of platters means that your actuator arm is bigger, and thus heavier. Having a heavy actuator arm means that you can't move it as fast, and thus your seek times suffer. Similarly, having a lot of platters means that it'll take a larger motor to spin them the same speed as it would for fewer platters, and thus your drive is going to be larger and produce more heat and noise at the same RPM.

In the end, it probably makes sense to use a limited multi-platter approach (maybe 2-4 platters), but keep RPM and platter density high. Additionally, unless you are talking about the really highend (say 15k RPM drives), it's currently more economical to raise the platter density and spindle speed rather than going for multiple actuators, as both approaches give an increase in bandwidth. There is also always the option of doing multi-disk raids if you need more bandwidth as well.

Nite_Hawk
 
Back
Top