Raid 0 on a single hard drive.., just a thought

[Jace]

I've been kicking around this idea a bit, thought I'd post to see what some other thoughts were. With the price of hard drives and Sdram so low, one way to stimulate the market and increase hard drive performance at the same time could be to do something similar to the following:

1 hard drive with 2 40GB platters that were capable of raid 0, giving you 40GB total storage. What I'm thinking of is having the manufacturer build the striping logic into the hard drive at a low enough level that it would be transparent to an Operating System. The drive could ship with a 16MB buffer and be pitched at media mavens or other niches of the computer enthusiast market, say power workstations.

I know that there are going to be many schools of thought on a concept like this, but I think that there would be quite a few adopters. Another thing that makes this technology more appealing is the ATA133 spec, and the upcoming Serial ATA specs that start to surpass current STR's by an appalling measure.

Currently an 80GB hard drive at tcwo.com costs $157. So say this 80/40GB drive costs $189. For $190 you have Raid 0 and only occupy the primary master, leaving you room for your DVD, CD-RW drives and you're a happy ordinary home user who didn't have to mess with a raid card or even OS level Raid settings and/or limitations (like XP home and Windows 9x). If you want more data protection, add another as secondary master and set NT Workstation/Server or 2K/XP Pro to mirror the drives and for only $378 you have 80GB in a Raid 0,1. (10) It would be cheaper to use the 80GB drives available today, but they would not be as flexible.

Closing thoughts; this idea is just to add another option to simplify getting more speed for some people. It's certainly not my thought to tout it as the be all end all, or as a complete replacement to existing technologies. Hope there are some good comments.

 

[AEboy128]

I never thought of that, It would be nice though.

 

[HappyPuppy]

The read/write heads would have to be independent for each platter. This would mean double the electronics and double the hardware. You would, in effect, have two harddrives in one housing.

Somebody correct me if I'm wrong.

 

[ModemMix]

No, your right. The only thing saved would be physical space, it would virtualy be two drives in one shell.

 

[LokutusofBorg]

IBM has had this independent head technology for some time. I don't know if it's currenly implemented in production units though. "Double the electronics, double the hardware" wouldn't be the case I don't think. More certainly, but not nearly double.

 

[FlipSide]

ditto puppy.

the question also on ide channel. raid 0 specifies 2 different channels, where will you get the speed increase on the same ide channel. The speed is still governed by the slowest component (bottleneck).

but you are right, the only way to improve things is having thought like this, and maybe not workable at first, but who knows a few a changes here and there then viola!!!

i always believed that success is 99% failure, persistence is the key, and of course a good funding source.hehehehe.

 

[marat]

<< I've been kicking around this idea a bit, thought I'd post to see what some other thoughts were. With the price of hard drives and Sdram so low, one way to stimulate the market and increase hard drive performance at the same time could be to do something similar to the following:

1 hard drive with 2 40GB platters that were capable of raid 0, giving you 40GB total storage. What I'm thinking of is having the manufacturer build the striping logic into the hard drive at a low enough level that it would be transparent to an Operating System. The drive could ship with a 16MB buffer and be pitched at media mavens or other niches of the computer enthusiast market, say power workstations.

I know that there are going to be many schools of thought on a concept like this, but I think that there would be quite a few adopters. Another thing that makes this technology more appealing is the ATA133 spec, and the upcoming Serial ATA specs that start to surpass current STR's by an appalling measure.

Currently an 80GB hard drive at tcwo.com costs $157. So say this 80/40GB drive costs $189. For $190 you have Raid 0 and only occupy the primary master, leaving you room for your DVD, CD-RW drives and you're a happy ordinary home user who didn't have to mess with a raid card or even OS level Raid settings and/or limitations (like XP home and Windows 9x). If you want more data protection, add another as secondary master and set NT Workstation/Server or 2K/XP Pro to mirror the drives and for only $378 you have 80GB in a Raid 0,1. (10) It would be cheaper to use the 80GB drives available today, but they would not be as flexible.

Closing thoughts; this idea is just to add another option to simplify getting more speed for some people. It's certainly not my thought to tout it as the be all end all, or as a complete replacement to existing technologies. Hope there are some good comments. >>




Can you explain the difference/advatage between harddrive with 3 platters (and thus 3 heads) @20GB per platter and you scheme (say 3 harddrives with 1 platter @20Gb each )? I mean either way, it is 3 heads and 3 platters @20Gb! Only ordinary solution is simpler, since heads are not independent. Sorry if I misundertood.

 

[boran]

<<
Can you explain the difference/advatage between harddrive with 3 platters (and thus 3 heads) @20GB per platter and you scheme (say 3 harddrives with 1 platter @20Gb each )? I mean either way, it is 3 heads and 3 platters @20Gb! Only ordinary solution is simpler, since heads are not independent. Sorry if I misundertood. >>




the difference here is that your 3 20 GB hd's will take up three ide channels and wont saturate any ide channel, (not even using 33 MB/sec so not even a third from the bandthwith) by combining 2 HD's into one, and applying raid 0 on those their speed will double and we'll get higher transfer speeds (true ultraATA as one might say ..)

and it saves space ...

 

[Scootin159]

There was a discussion in HT about a week ago on this very topic. I think the end result was that it would essentially be 2 drives in one housing like happypuppy said. Even though there wouldn't be exactly 2x the electronics, it will actually be more expensive due to the space contraints for the various components (you're basically making a 1/4 height drive - standard drives are 1/2 height).

 

[Jace]

Well, looks like I'm not the only one interested in this idea. Perhaps the hdd manufacturers will consider this, unless they have some other ideas for boosting speed of single hard drives. Increasing data density and spindle speed seem to be their main focus, other than the recent addition of an 8mb buffer on Western Digital's new hdd which was very cool.

In my first post I used "Raid 0" to help describe the thought process, when in actuality, the hdd manufacturer wouldn't need to follow exact RAID specs, such as using 2 channels for striping. Rather, they would need to come up with (proprietary, yet transparent outside the confines of the hdd) independent head mechanisms and controller logic which write data to the multiple platters simultaneously in a striping mode.

Maybe there will be some other breakthrough, like capacious MRAM drives that will make this sort of discussion moot.. like when I'm 90 ;-)

 

[thomsbrain]

I thought that was the way that all hard drives work to begin with. I mean, most drives are multi-platter, and they write from the outside in on all platters at once. They don't fill one and then start the next. That's why hard drives get slower when they get full, instead of always getting slower and faster the more stuff you put on them.

Correct me if I'm wrong, but that was my understanding.

 

[SCSIRAID]

The problem with the concept is that you would need two different actuators (servo and arms and heads and preamps and data separators etc etc). This concept has been considered for quite a while now but has always come back as cost prohibitive and insufficient interest. It is certainly possible though. The reason you need a second head stack is that the servo positioning information is embedded on the data tracks on a per head basis. When you are 'on track' for say head 1, head 3 could be quite a bit 'off track' due to stack tilt, thermal expn etc. Currently there is only one read/write channel so you would need more channels to transfer data in parallel.

 

[darth maul]

<< I thought that was the way that all hard drives work to begin with. I mean, most drives are multi-platter, and they write from the outside in on all platters at once. They don't fill one and then start the next. That's why hard drives get slower when they get full, instead of always getting slower and faster the more stuff you put on them. >>



Well if that were true, wouldn't a drive based on the same technology (same platter sizes, motor speeds, heads, etc) but coming with twice the number of platters be two times as fast? Like drives based on 20GB platters, where on drive is just a 20GB drive, and the other drive is a 40GB drive. If the 40GB is writing to all platters at once, wouldn't it be twice as fast as the drive with a single 20GB platter?

 

[Scootin159]

Yeah, I could see the problem with having each arm operate independantly, but why not put two entirely seperate read/write head assemblies in there. This should effectivly cut seek times in half as well as doubling the sustained data rate. I do see the potential cost issue (two assemblies = 2x the cost & 2x the electronics), but even if you figure in the initial cost being 2x as much, I would pay $120 for a 20GB version of this drive. I know they will argue that it is a small market, but if they moved it into SCSI territory they will always find a market for extreme high-end devices.

 

[Pariah]

A few years back Conner developed a drive that could read/write from multiple heads at once. The head did not move independently, it just had the ability to read and write from 2 heads at once. The company was subsequently bought out by Seagate and I believe they were the ones that actually released the drive. Problem was, it was so complex and the performance gains weren't that great so the technology was cancelled.

Due to the way drives' mechanics work, I don't think it is possible to make the heads in a single actuator move independently. Basically the movements of the heads are controlled by a powerful magnet with a voice coil attached to it that is fed a current moving the entire assembly. Clearly there would a problem trying to stack powerful magnets on top of each other and trying to get them to move independently. You could try multiple actuators but that would require an entirely new hard drive form factor to fit, and the benefits just don't seem to be there. It's cheaper and easier to produce 2 drives than trying to squeeze 2 into 1.

 

[thomsbrain]

<<

<< I thought that was the way that all hard drives work to begin with. I mean, most drives are multi-platter, and they write from the outside in on all platters at once. They don't fill one and then start the next. That's why hard drives get slower when they get full, instead of always getting slower and faster the more stuff you put on them. >>



Well if that were true, wouldn't a drive based on the same technology (same platter sizes, motor speeds, heads, etc) but coming with twice the number of platters be two times as fast? Like drives based on 20GB platters, where on drive is just a 20GB drive, and the other drive is a 40GB drive. If the 40GB is writing to all platters at once, wouldn't it be twice as fast as the drive with a single 20GB platter? >>



ah yes, exellent point. so do hard disks get slower and faster as they fill? do those hard disk speed tests that show inner and outer speeds just find the correct point in the drive to do the test?