RAIDrive

All things storage here: hard drives, DVD RW drives, little wicker baskets.

Moderators: morphine, Steel

Postposted on Tue Feb 19, 2002 5:24 pm

Ok, I have no clue how hard drives operate. Honestly, it's the least interesting part of my system, so I never bothered to explore the technology. But I've got this idea in my head that just won't go away.
Doubling up on everything seems to be really popular these days in computer technology. We've got dual-core CPUs, hyperthreaded cores, dual vertex shaders, SLI (ok, that one is pretty much dead now), etc.
So why can't a HDD manufacturer implement a single hard drive that functions as a Raid 0 array. Don't HDDs have multiple platters to begin with? Couldn't these platters be assigned to striping duty? I would think this would be an ideal solution to better HDD performance. A fast HDD system in a single unit (for better performance and lower costs).
Somebody explain the problems with this to me.
Trident Troll
 

Postposted on Tue Feb 19, 2002 5:56 pm

One immediate problem with that idea is that you need someone whose motherboard is RAID capable, else you need some sort of logic inside the HDD to work the RAID...

Computer manufacturers have a habit of going for the lowest common denominator, which is the basic motherboard with no on board anything... just the basic essentials in the slowest forms. HDD manufacturers are making drives that will still work with ATA33 for Christ's Sakes!!!

Another problem is that if the HDD logic was implemented, it would lower the Mean Time Before Failure (MTBF), which is the standard that all drives are measured against, and the reason why IBM drives are getting a hammering recently.

Oh yeah, and finally, it wouldn't work :smile:

RAID 0 relies on the two disk heads being able to read and write both two bits at a time in order to speed up the data handling capabilities of the system... In order for RAID 0 to work, two disk heads would be needed, which would bump the cost up even more!

RAID 1 is totally not possible on a single drive, because it is a data protection system. RAID 1 duplicates the data across two drives, so if one fails, the other still has the data to recover. If you use a single drive and it fails, you're knackered...

Because RAID 1 would be impossible, so would RAID 0+1, which does RAID 0 on two drives, and copies these drives onto two separate disks...

Sorry about the length and disorder of the post, but hey, you get what you pay for :smile:,
IntelMole
IntelMole
Grand Gerbil Poohbah
 
Posts: 3529
Joined: Sat Dec 29, 2001 7:00 pm
Location: The nearest pub

Postposted on Tue Feb 19, 2002 9:01 pm

Take a spin over here to learn more than you ever wanted to know about how hard drives work.

Raid 0 on a single drive sounds interesting but there's a ton of little techical issues I can't think of right now why it won't work (or work well). Things like head alignment, increased complexity and increased chance for failure. (I guess I could think of a few... :wink: )
Steel
Global Moderator
Gold subscriber
 
 
Posts: 2326
Joined: Wed Dec 26, 2001 7:00 pm

Postposted on Tue Feb 19, 2002 10:11 pm

To add to that, I have seen the workings of a drive, and it would be very difficult to squeeze another head/control mechanism in there. It's pretty bad as it is now, and to split up the separate heads would be nasty. If anyone has a pic of the inside of a drive, it would explain things better...

Also, ditto to the above comments, which I wouldn't have thought of on my own...
highlandr
Gerbil Elite
 
Posts: 542
Joined: Thu Dec 27, 2001 7:00 pm
Location: Somewhere in downstate IL

Postposted on Tue Feb 19, 2002 10:58 pm

The link to Storage Review has some pictures.
Steel
Global Moderator
Gold subscriber
 
 
Posts: 2326
Joined: Wed Dec 26, 2001 7:00 pm

Postposted on Wed Feb 20, 2002 3:13 am

On 2002-02-19 16:56, IntelMole wrote:
RAID 0 relies on the two disk heads being able to read and write both two bits at a time in order to speed up the data handling capabilities of the system... In order for RAID 0 to work, two disk heads would be needed, which would bump the cost up even more!


Steel wrote:
Raid 0 on a single drive sounds interesting but there's a ton of little techical issues I can't think of right now why it won't work (or work well). Things like head alignment, increased complexity and increased chance for failure.


highlandr wrote:
To add to that, I have seen the workings of a drive, and it would be very difficult to squeeze another head/control mechanism in there. It's pretty bad as it is now, and to split up the separate heads would be nasty. If anyone has a pic of the inside of a drive, it would explain things better...



Yeah, well, that's what engineers are for...
Trident Troll
 

Postposted on Wed Feb 20, 2002 9:22 am

Well to add an answer to what you asked
It is already a stripe set in the hd
On each plater, the read/write head writes 1 bit to each plater in parrallel
Say you have 4 platers with 2 face, then the hard disk will write 8 bit at a time.
It doesn't reduce the mtbf since it is the same mechanics that works for all of this compared to a raid-0 where you have 2 head, 2 spindle motors and 2 electronics working at the same time.
Nuclear
Gerbil
 
Posts: 15
Joined: Tue Jan 29, 2002 7:00 pm
Location: Quebec, Canada

Postposted on Wed Feb 20, 2002 9:28 am

Um, no, that's not how it works. Multi-platter hard drives would be much faster if it did. When reading or writing only one head is active at a time.
Steel
Global Moderator
Gold subscriber
 
 
Posts: 2326
Joined: Wed Dec 26, 2001 7:00 pm

Postposted on Wed Feb 20, 2002 9:36 am

On 2002-02-20 02:13, Trident Troll wrote:

Yeah, well, that's what engineers are for...


Well why did you ask the question in the first place if you already believe engineers can accomplish anything :wink:. It's probably not impossible to do RAID 0 in a single drive, just technically difficult and cost prohibitive.
Steel
Global Moderator
Gold subscriber
 
 
Posts: 2326
Joined: Wed Dec 26, 2001 7:00 pm

Postposted on Thu Feb 21, 2002 1:03 am

Because I was hoping for a little more creativity on the part of my fellow TR forumgoers. Instead, all I got was either 1) It's impossible or 2) It's impractical.
I don't care if it's impractical. That is what the engineers are for. As for the other argument, I wasn't provided enough evidence to prove to me that it is impossible.

I was just thinking that maybe some of you would consider the possibilities. If not a true RAID system, then perhaps interleaved heads (thx for that SR link, btw). Who knows, perhaps something like this...

Image

If the platter diameter was significantly decreased, you could fit such an assembly in a regular 3.5 drive case (to say nothing of the myriad possibilities that IBM's microdrive carries with it). This should also result in lower seek/access times. And the heads could alternate sectors for any given data set. This would be a sort of 'Virtual RAID 0'. I can think of plenty of other internal configurations, though.
Yes, a new type of HDD architecture would be an expensive undertaking for a manufacturer. But innovation almost always carries risks with it. And, if successful, the risk-taker ends up on top, and everyone else is playing catch-up.
Trident Troll
 

Postposted on Thu Feb 21, 2002 5:45 am

Of course such a thing is feasible. But what you have to remember is that by having two seperate heads reading the platters you're effectively halving the amount of space on the disk so in the long run it's probably best to buy two HDD's anyway.

Space of course is a concern but with all the various HDD's being developed it won't be an issue in a few years. Besides, problems like these keep us in a job! Do we really want to make ourselves redundant?



<font size=-1>[ This Message was edited by: Keldorn on 2002-02-21 04:46 ]</font>
Keldorn
Gerbil In Training
 
Posts: 1
Joined: Wed Feb 20, 2002 7:00 pm

Postposted on Thu Feb 21, 2002 3:51 pm

Somebody explain the problems with this to me.
I thought you wanted reasons why this hasn't been done by the HD manufacturers, sorry if I misunderstood.

ImageThis was actually tried once - I can't remember the manufacturer - but I think they did this to improve access times more than transfer rate.
Steel
Global Moderator
Gold subscriber
 
 
Posts: 2326
Joined: Wed Dec 26, 2001 7:00 pm

Postposted on Thu Apr 04, 2002 10:55 am

If you're still interested in this topic, I found a link to some info on a drive this was actually done in:
http://www.redhill.net.au/d-o.html#2hp

Found it in this thread at Storage Review -
http://forums.storagereview.net/viewtopic.php?t=2344
- where this kind of question comes up about once a month.

Edit: corrected the URL for the 2HP.
Steel
Global Moderator
Gold subscriber
 
 
Posts: 2326
Joined: Wed Dec 26, 2001 7:00 pm


Return to Storage

Who is online

Users browsing this forum: CityEater and 8 guests