How is a partition split?

[Zak33]

Hi guys,

When you partition a drive (say equal shares) HOW IS THE DRIVE DIVIDED? I do'nt mean software-wise. I mean , physically. Is there any logic to the division? If it was, say a 3 platter hard drive, does each partition get 1 1/2 platters?

Or, more sensibly, does one partition get the outter edge of all the platters, making it the faster partition?

Or is it totally totally random?

I've been pondering this for a long long time, and I just can't find the answer. If it was a Barracuda 160gig SATA drive, which has only one platter, is it divided into two with one partition getting the outter part? Or does one partition get a cake-slice-shaped wedge?

Reason I ask is that I'm gonne build a SFF, and there'll be a Raptor boot drive, but the second drive will be a larger unit, and I am considering partitions, one for games and one for data. I currently use 3 seperate drives (Boot/Games/Data)

Any advice or experience gratefully received

[aidanjt]

Most drives start from sector 0 at the outer track, so partitions from there are generally faster, more sectors allocated to the partition the further in it that partition ranges to. So it's more like a 2D onion, and you can group the rings into partitions.

[kalniel]

Yup, but the question about platters is interesting - the outside track on a second platter will be faster than the inside track on a first platter.

For simplicity I would guess it follows a linear arrangement so as you work your way along the volume you actually get an increase in speed when you jump to the second platter -ie it basically follows a spiral in from the outside for each platter. This is why I set my partitions to be roughly equal to platter size.

[Gerrard]

I belive the heads for the different platters are all fixed together (ie all move together). Therefore, it would seem more logical to read all the outer tracks consecutively (or even simultaneously maybe?) from each platter before moving the heads in to the next track. If this is the case, then the partition would be split along the outer third-ish (due to zone recording there will be more sectors per track on the outside) of all the platters. The first partition would also be quicker than the second due to being on the outside of the disk.

Not certain on all of this, just seemed the more practical way of doing it....

[kalniel]

That's true, they are all fixed, so in effect the data is striped across all platters?

[EtheAv8r]

When you partition a drive, the first partition takes the first n cylinders from the outside edge working inwards (n depending on the size of the partition), the next partition created takes the next n available cylinders working towards the middle.

Outside tacks/cylinders are faster, therefore the first partition will have the best performance.... untill that is you also start accessing data on the second partition as at the same time, then the heads will be busy working on both partitions.

[kalniel]

When you partition a drive, the first partition takes the first n cylinders from the outside edge working inwards (n depending on the size of the partition), the next partition created takes the next n available cylinders working towards the middle.

But what about a two platter disk?

[kidzer]

If memory serves, a Cylinder in this context is a particular track on every platter, so the first n cyclinders would take up n tracks on each platter, I think.

[Zak33]

this is good....you lot are quite cool at advice for this kinda thing, and it's why I asked.

I'm gonna partition the drive and put the games in the first partition and the normal non speed critical data in the outer part.

I cant use HD Tachometer to test them cos it looks at the whoel drive as an item ( I think) but I wonder if there's another way to test.

Nero?

[badass]

If memory serves, a Cylinder in this context is a particular track on every

Yep

platter, so the first n cyclinders would take up n tracks on each platter, I think.

Dunno about that one myself but it sounds right

[Defenestration]

I'm gonna partition the drive and put the games in the first partition and the normal non speed critical data in the outer part.

You've mis-understood, slightly. The outer part of the disk is faster than the inner part of the disk. When partitions are created they are first placed at the outer part of the disk, with subsequently created partitions being placed spiralling inwards. Therefore, the first partition created will be outer-most, and hence fastest. The second partition created will be slower than the 1st, but faster than the 3rd partition. The 3rd partition will be slower than the 2nd, but faster than the 4th. And so on......

So, as you say, the first partition is where you want to put the games, and this will reside on the outer part of the disk. The normal non speed critical data should then be placed on the 2nd partition, which will be nearer the inner part of the disk.

HTH

[Zak33]

You've mis-understood, slightly. The outer part of the disk is faster than the inner part of the disk. When partitions are created they are first placed at the outer part of the disk, with subsequently created partitions being placed spiralling inwards. Therefore, the first partition created will be outer-most, and hence fastest. The second partition created will be slower than the 1st, but faster than the 3rd partition. The 3rd partition will be slower than the 2nd, but faster than the 4th. And so on......

So, as you say, the first partition is where you want to put the games, and this will reside on the outer part of the disk. The normal non speed critical data should then be placed on the 2nd partition, which will be nearer the inner part of the disk.

HTH

Sorry....I did understand.....I just worded it like a plum. I was thinking second partition....dunno where outer came from

cheers though



Once I've done those partitions I will try to ascertain speed differences.

[EtheAv8r]

But what about a two platter disk?

Sorry, I missed this.... a Cylinder.... if you like is a vertical set of tracks.... ie Cylinder 1 is track 1 on all platters, so all the outside tracks on all platters are assigned to the partition. Does that help?

[kalniel]

Excellent. Yes

I was at a conference a while back and some guy was presenting a poster about his PhD to do with speeding up data retrieval for hard disks and I asked him the same sort of questions. He couldn't answer, which doesn't bode well for his viva..