Which disks are faster?

humanoid

I have some questions:
Why WD raptor has only 37GB (74GB)? It is because of the 10k rpm? Which disks are faster, the bigger or the smaller one?
If we have one 200GB disk and one 40GB disk, which would be faster? For example, the disks are the same, same mechanism everything the same, same manufacturer, same year, but 200GB disk has 80GB per platter and the 40GB has 40GB per platter. Which woul be faster?

Why are SCSI disks also "small" max 80GB or something like that and also the Raptor has only 36 (74) GB. Is this becase of the speed? Why is that a problem, if a disk rotates 10k per minute why can't it be 160GB disk? Wouldn't the reading and writing be faster, because of the bigger density of the data?

Example:
160GB disk:
_ . _ . _ . _ . _ . _ . _ . _ . _ . _ . _

40GB disk:
_ .. _ .. _ .. _ .. _ .. _ .. _

_ = data
. = 0

The disk head can access the data faster, because of the density, but on the 40GB the density is much smaller so the disk needs to rotate more to access the same data. Is this correct?

Why aren't 200GB SCSI disks on the market?

10× for help

Tex

The larger raptors are faster. They are the second generation raptors. They made inprovements in a number of areas on the next gen raptors just like they generally do in each successive generation of any line of drives. Each new generation of scsi drives add 30 to 40 percent in STR. Usually helped greatly by increased aureal density made possible by improved manufacturing techniques and materials

And scsi drives are not limited to 80gb. The 73gb ones have been around like 3 or 4 years. They have had 147gb ones a couple years and the 300gb ones are shipping. But you lose the beauty of scsi with one large drive. You want multiple small drives. Scsi can read and write to multiple drives at once so your goal is to divide the I-O between the drives for much better performance. I wish more people could grasp that concept.

Tex

RWB

But why are they more limited by capacity than normal IDE?

Tex

I think it has to do with how dense you can get the data per platter and still be read reliably at 15,000 rpm. The goal with scsi is speed so when you compromise the speed its a tradeoff that doesnt make sense.

Its easier to read at 7200 rpm then 15,000 rpm so the larger capacities hit ide first. It just takes scsi a little longer to get there. But we have only had 300gb ide drives a relatively short time also if you think about it.

The huge differance is really in price as those 300gb scsi drives are very expensive also.

I use big arrays of 18gb and 36gb scsi drives. And they are pretty cheap even for very fast drives. I still use the cheap high capacity ide drives in all my servers too. I mean there is no reason in the world to stick 40gb of MP3's and that much porn onto fast scsi drives when any cheap 120gb ide will handle dat like that just fine. I love scsi and have multiple dual cpu servers with high end scsi raid controllers on a gigabit network but I still back the scsi arrays up and keep all my casual data like movies and mp3's etc.. on cheap ide drives and keep them backed up to other cheap ide drives.

Remember any single drive can only handle so many I-O's per second. So you want more smaller drives rather then one huge one. The smaller drives handle roughly the same I-O's per second but with lets say eight 36gb drives rather then one 300gb drive you can handle a much heavier load.

Tex

Straight_Man

Yes, partly that and that the platter material absolutely has to not distort at all at higher rotation rate-- if the media ever gets out of a pure flat spin, not only does data storage get unreliable but heads can crash on it. So, the faster HDs have more rigid media, bigger rotational and stepper motors, and heavier bearing structures than IDE drives which rotate slower. Some of the biggest Enterprise grade HDs these days are Fiber Channel HDs, but they are not priced for consumers by any means.

BTW, the HDs you see with the sizes of first gen, are mostly single platter HDs, using both sides of platter or even just one side of a platter for small capacity drives.

As to which would be faster in reality:

The further the read\write head has to move over platter before it starts writing, the more time it takes before wile write or read starts. NOW, the faster the HD spins, the harder it is to get the positioning exactly right. The more dense the storage, the harder it is to get the extreme precision needed to have head write exactly where needed. So, if you go faster and denser at once, especially with newer technology, you get a situation where if the movements are not so dang close to perfect that the embedded controller card on HD has no need to try and reseek to get head exactly where it needs to be then you get some movement back and forth of head as it zeroes in on exact right position. OR, you have to go less dense as you spin drive faster.

In theory, if you get a perfect drive and perfect mechanical movement in drive, faster spin is faster as it takes less time to move to point to start to write and actually write onto spinning platter. BUT, in reality there are only so many things that can be dond without making HD bigger form factor also and getting 99% reliable media 99% of the time is hard enough as it is. So, some folks who use latest tech tell the controller that handles the channel to write verify. This means that while write verify is on, that it takes about 2.25 to 2.5 times longer than with wirte verify off to get a cross check of data write. Head has to write, stop, move to start point, read, adn if errors after a checksum calc is performed, REWRITE. with first gen drives, I would definitely write-verify for a while at least, if not design a HD to write-verify by default.

Second gen should be faster and more reliable if the two HDs were same size. But, if HD that is much bigger does the storage on same number of platter surfaces, then the starting, positioning, and stopping have to be calced and performed faster and more reliably or the storage of file info gets radically unreliable. So, it tends to be a drive that will be in write-verify mode if the data is real valuable. AND it will take longer to write-verify than for the smaller drive to just write.

The two almost balance out, as the bigger HDs run storage onto more than one platter while the smaller ones run a single platter or even a single side of one platter for storage only if real small. For BIG to huge files, the bigger HD should be faster, for many small files it could well be same or slower. The less the head has to be moved to a new start point or back to start to read data to write verify aginst what should have been written, the better and more really effective the drive is. For normal use, if you need to not have real huge files bigger than 1\3 of the HD capacity written and many smaller ones, you are better off to have more smaller HDs than one huge one. For HUGE or even very large files, the bigger one will be better generally.

What Tex said is right, totally. I simply read from what was asked all together and the way it was asked that a further explanation might be in order, in more plain english than pure techese. HTH to show why bigger and faster can sometimes be better and sometimes NOT be better. Much more than just engineering-designed capacity and speed designs aggregate into what you get for net real performance figures.

humanoid

Thank you very much for your answers

For the SCSI:
If one has the money to buy 10 x 300GB disk and 10 x 30GB disks, which would be faster?
I presume that the 300GB combination would be faster.