Sign in with
Sign up | Sign in
Your question

Raid harddrive partitions based on platter size

Last response: in Storage
Share
Anonymous
a b G Storage
July 7, 2010 2:38:31 AM

Hello,
I have in the past done a raid with a single hard drive, the harddrive was 1tb, with 4 platters each 250gb, and was able to raid the 1tb down to 1 250gb Raid harddrive, where the 4 platters essentially acted as its own harddrive with 2 heads per platter, i wanted to know if anyone has done this and what the actual speed benefits would be.
a c 415 G Storage
July 7, 2010 3:05:54 AM

Hard drives do not write data one platter after another like a stack of vinyl records. They write data by first filling all of the outermost tracks on both sides of all platters, then the heads move in 1 track and the next set of tracks on both sides of all platters is written, etc.
m
0
l
Anonymous
a b G Storage
July 7, 2010 3:51:51 AM

Well, your not understanding what i am saying,
Hypothetical
2 Harddrives 500gb Each
Each Harddrive has 2 platters 250gb each
so you create a raid partition of 250gb, then create another 250gb Raid Partition,
if you know the platter size ie, Full size Divided by Number of Platters, then you create a partition of the size of each platter, its going to use them incrementally
so you raid the 2 harddrives together, Platter 1 from each harddrive is its own raid Partition
PLatter 2 from each harddrive is its own Raid Partition, so you now have 2 harddrives with 4 partitions in a Raid 0, ok

so think about it, you have HD1 PLatter 1, and HD2 platter 1 storing the same information with its own 2 heads reading the information from 2 separate drives, faster,
RAID will use each drive to store the same information so that you can read it from 2 places rather than 1, doubling the read threads and doubling the speed.

i know it is possible, i have done it on servers before i was just wondering if anyone has done it for a more enthusiast application, to see the benefits.

The harddrive is written to however you tell it to write to it. Striped, mirrored, by cluster size ect..

when you are creating a Raid Partition you select the size of partition you want to create, if you are using 2 identical drives with the same platter size, then you figure out how many platters you have and what size they are in Kb then create your raid partition that size, you create a separate raid partition for each platter depending on the size, it will use it incrementally so if you enter a partition size of 250gb and each platter is 250gb then thats all its going to use, because you said to use from 0 to 250000mb first, all of witch is contained on the first platter, then if you create another portition thats 250 then it just so happens to land on the second partition, this isnt a setting thing its a math thing if you know the platter size then you can get the RAID partitions to land just right so that Raid HD1 is on the First platter of each drive, then RAID HD partition 2 will be on platter 2, is this a better explanation?
m
0
l
Related resources
a c 415 G Storage
July 7, 2010 4:45:04 PM

I'm sorry, but your understanding of how hard drives work is wrong. First of all, the heads for each platter do not move independently of each other - they're all fixed to the same actuator. When the heads for one platter move to the outermost track, all the heads move to the outermost track.

In the past there were a few large industrial hard drives used in mainframe computers that had multiple access arms and which could have heads positioned to different tracks, but no modern 5-1/4" or smaller hard drive has ever had that capability.

All of the tracks that can be accessed without having to move the head assembly are called a "cylinder". Data is written to the drive on a cylinder-by-cylinder basis, one track at a time starting with the outermost cylinder and moving toward the innermost cylinder. Logical block "0" of a drive is located on the first track of the first platter of the outermost cylinder, and as you increment through logical block numbers the heads move inward one "notch" after all the tracks on the outermost cylinder are filled. This strategy is used because it reduces the number of times the heads have to be moved while reading/writing sequentially and therefore increases performance.

Note that although there is a separate head for each surface of each platter, the drive can only transfer data through one head at a time. This is because differential thermal expansion of the platters combined with the extremely tight tolerances in modern drives mean that it's impossible to guarantee that all heads can be aligned with all the tracks at the same time.
m
0
l
a b G Storage
July 7, 2010 4:50:30 PM

Just vouching for what sminlal wrote.
m
0
l
!