Raid controllers are fine for data protection, but specially in windows, I do not see so much performance increase using raid in 1 0 or 5.
Why the Hard Drives companies does not make a hard drive with two or more reading heads? it is that mechanical and logical possible? could be "Full Duplex" drives. I think so.

I always thought the same, clarifying before someone flames, we are talking about two heads over the same plate and side (meaning 4 heads each plate)

But, we have to also take into account they'll be twice as heavy, hence slower to seek since you deal with more mass and inertia to start and end the movement.

With two heads per platter side and all heads capable of reading/writing bits in parallel to/from all platters simultaneously...we might have something that can compete with up and coming technologies in the future. I had this idea over 10 years ago...still waiting.

"They" do have such drives. They are just not for sale. And if you try and build them, expect being nuked any time. Seriously, this is a part where military technology is way ahead. Magnetic drives with just the exterior part of the magnetic disk data-able, multiple independent heads on different positions around the disk, working in a concert of read-write with awesome speed and minimal seek times.

More heads dont mean slower access and higher enertia, because they are independent, ou gain twice de read and write capability and faster seek times, because you can use the head that is nearer of the target point. It is a great ideia, but I think it should be included in the harddrives and not the controlers, although it could make them cheaper, but the protocol should be standard to allow multiple heads in diferent discs.

The 2 heads per platter was thought of 30 years ago -
Mainframe disk used to have 2 head arrays for speed. Halved the latency. Very expensive/heavy and prone to failure if moved, was really only useful for big arrays that sat in one place forever.

...read/write head, one for each plater side. So a drive with 2 plates will have 4 heads. But, they are all moving together on a SINGLE arm, so they are no independent to perform simultaneous operations. It seems like enabling a separate arm for each head would greatly raise the complication and price of the hard drive. Maybe even to the point of making it none cost effective as it could pass the price of an additional drive or more (so you could gain a similar performance in a RAID setup, but with the benefit of the additional storage space).

This is really a great product! Apart of being a raid controller it includes a 56k modem and its buzzer!

How would more heads read faster? I'm thinking of a record player with 2 or more neddles. Or a CD/DVD with 3 lasers, that helps with skipping. But reading the same data faster?

I would also think more moving parts would lend itself to failure, cost, and power.

I am by no means an expert, so forgive my ignorance.

I'm no engineer, but I'm pretty sure if you increase the heads per platter, you would raise the cost of the drive a fair bit. Not only is there added complexity and components in the actual mechanics, but I'm sure the actual controller logic would be pretty tough to develop too. Not to mention the increased chance of failure due to this added complexity.

Considering how drive development is going, my guess is the manufacturers believe it is more cost effective to get speed increases (although minor) by increasing the density of the platters. This has the added benefit of greater top end storage, and cheaper costs for your lower models. Not as fast, but more profitable.

Hopefully someone with more knowledge will come along and shed some light on this though, because I do agree that HDD speeds are getting ridiculously slow in comparison to the rest of the PC.

There were multibeam cd-rom drives - same concept - parallel access to the disc. I bought 2 of them for about $90 each or so and they were super fast!
But since the average consumer doesn't want to spend about 75% more they stopped making them - which sucks!

This is cool and all in the short term but all drives will soon be Solid State drives or IoDrives and we will have no more spinning disk technology.

PC Computing had an article in March 1990 which discussed optical memory. This optical memory would have no moving parts, be no larger than a sugar cube, and have a theoretical limit of 1.25 Terabytes. It uses several lasers aimed into the cube to change the color of molecules which represent data. This was 17½ years ago. The article mentioned financing by the Air Force. Why such a storage type was never produced, I don't know. Perhaps the military only wanted it for itself. Perhaps it was never commercially feasible or maybe the memory itself was not stable.

I'm surprised someone would say RAID wouldn't improve performance of windows.

Why wouldn't increasing free space & read speed w/ a RAID 0 improve performance?

RAID 0 w/ two HDD's doubles read speed. On my RAID 0, if I added a 3rd HDD, it would go even faster; of course that only helps w/ loading large programs [huh, like windows] and games, etc.

RAID 1 doesn't help with speed, actually it slows it slightly, but you've got two identical drives in case one fails. Everyone could see the benefit in that, even if they don't think they need it.

I'm using a RAID 0 & a RAID 1 array, both w/ 2 SATAII drives each. Now that I've actually used it, I just can't see myself not using RAID.

Just because people don't see a need to use it themselves, doesn't mean it's not useful.

Having GTA: SA installed on my RAID 0 cuts the loading times in half, I find that pretty darn useful.

Regardless, the point of the above article is mainly for new tech. aimed at high-end workstations & servers.

If you're running a mission critical server, not having RAID is stupid.

The reason they stopped making the multi-beam CD-ROM drives surely must be that they were unreliable. Otherwise the class action lawsuit would not have gotten my friend and I each a replacement drive after a short use. Our replacements failed within a year. Great idea though...they were screaming fast.

As far as hard drives go, OK, forget the multiple head per side if reliability worries you. Just implement the parallel data streams idea so each head is routed part of the data stream. Just change the drive circuitry and firmware. The physical sectors would be mapped across multiple platters, but the firmware would hide this fact from you. Better data integrity (and manufacturer profits) would come from offering a RAID like option where usage of one head would spread correction code. Add two such drives to a real RAID controller and you would have unparalleled speed and data integrity.

Many people who get the idea of increasing their storage performance by using two drives in RAID 0, are not aware of the danger with such a setup. A RAID 0 setup would be ruined if either drive fails, so the chances for such a person to loose their whole data is 2 times of the chance for a single drive failure.
i.e. If a specific WD Raptor has a failure chance of 5% per year, then a RAID 0 setup with two of those drives would have a failure chance of 10% per year.

The only reasonable solution (by my standards) would be to add a 3rd drive for backup (large enough for the combined size of the two drives in RAID 0, but could be slower).
Tom

I would hazard a guess that if you had 2 heads per platter they could possibly interfere with each other electromagnetically during write cycles since the platter is metal or glass coated with some sort of metal foil/coating/paint. It would be interesting to hear if someone has tested something like that.

On the discussion about multiheads - a historical recap:
Mainframes had 2 sets of heads per platter set on each side to reduce latency (a data seek would take half as long reducing access time).
In Tandems (high reliability mainframe) they also had 2 heads, which were in constant motion back and forth over the platter to increase the chance of requested data being nearby.
When combined with multiple data paths to the disk, and on board processors, you could get sustained 80MB/sec transfers from disk packs WITHOUT any kind of RAID - and this was 30 years ago.
The drawback was cost - very expensive disks. These days mainframe disk is just huge arrays of cheapish RAID disk.

At these densities and nanometer sizes its too hard for the two heads to read and write data in a fashion that would enable them to interoperate. We are talking about quantum effects with electonic spins, kinda a hard for two different heads to match each other precisely enough.

Practically speaking, you get better throughput between multiple drives rather than multiple heads because each drive has its own personal datapath when decoding the signals.

By the way, most drives only use a single platter. Add more platters and it slows down the drive due to mass effects. Even that 1TB drive is a single platter.

Finally, manufacturing these things is a delicate process with each step very controlled. It takes a long time to get yields up. Add too many steps and you'll never be able to make a drive.

The military drives are built to take higher levels of shock and such, not go fast. More heads only increase the contact points for damage. That guy must not understand what he's working with.

not quite
the tb drives use like 4 or 5 platters
i think the most gb per platter right now is what 250?