Imagine a HDD (500GB) with 1 platter, 1 head. It has an average W/R speed of say 100MB/s Now a hdd (1000GB) with 1 platter, 1 head, it will still have and average os say 100MB/s. Thats cos with both drives. any any one time, only one head is active. Now imagine the second drive, where both heads work together to read write data. Just like RAID0 but within one drive. I dont see why it cannot happen, as everything required is available. The hardware is already in place, all that needs to be done is the IDE be redesigned to allow RAID0 like behaviour. So right now, at peak, drives with single head sine platter can do about 200MB/s, and if it was a dual platter, quad head....you get the idea :nerd:
this a hard drive head stack first two pair goes for first platter and so on,one above,one beneath .data is written in cylinders,imagine a cake cut a piece of it,data starts on first layer,then next beneath it ,second layer the second beneath... so,a platter always has 2 heads reading/writing,firmware says when it applies power to each one, for each task needed. IMHO ,increasing platter density will increase write/read speed,but also needs better faster algorithms for on-the-fly ECC. i may be wrong,tough.
ok so if say 800 MB needs to be written, and each head can write 100MB a seconds, 8 heads, together, can write 800MB/s Right now the technology only allows one head to operate at any given time. Thats why if you have a 500GB HDD 1platter, 1 head, and and say that takes 30 minutes to fill, a 1TB with one platter, 2 heads will take 60 minutes to fill
Yeap. When chip costs go down, we wont need magnetic disc´s or moving parts at all. Soon everything will be digital, specially as internet speeds go up, we need harddrives that can keep up. Old HDD´s are soon to be obsolete!?
Not really, the internet speed depends on CPU and RAM. Then it flushes to the HDD. It doesn't make any difference.
First bottleneck to be hit is going to be mechanical HDD's though. I've seen a 10Gbit connection being bottlenecked by a HDD before... Everything mechanical in the PC world its performance increases only gradually compares to the rest.
Sorry... but it's going to be a looong ways off yet, mechanical drives still a healthy lifespan, especially for server storage.
but cost wise, if you look at it, they wont need to add any more physical stuff to the drive, just update the conroller to use all heads at the same time think raid0 in one physical drive
Wouldn't that cause massive fragmentation if there were 4 platters and 8 heads all writing the same file to different platters ? Wouldn't it increase fragmentation by a factor of 8 ?
it would be o different then a ssd writing to say 8 nand chips technically it would be fragmentation (depends on how you look at it) so instead of it taking 8 times more to read it off a 'current' drive in a linear fashion, it will be read back 8 times faster if the file was split across the 8 parts and each head reading each part concurrently
I am not an electrical engineer, but I'd imagine that reading and writing to all the heads in parallel would require a dedicated signal path for each head including amplifiers, decoders, signal processors, etc. So it could be a cost thing, but I did some searching and found another discussion on the topic: Storagereview.com It seems that we've reached the point where track densities are so high that it is very impractical synchronize reading from multiple heads at once. It is technically possible, but keeping all heads aligned simultaneously under all temperature conditions is difficult with current (aka cheap) actuator technology. HDD's are all about price per gig, speed is now solid-state territory. If you need fast mass storage, its RAID or nothing.
Multi-head technology actually does exist and is used, but does not provide a huge speed benefit; not only does it significantly increase the power consumption of the drive due to the required processing power, but it would increase the points of failure on the drive, along with the chances for data corruption. You would have to increase the space around each sector, thus losing data density. The reason it doesn't provide a speed increase is because you have to work the drive more slowly to make sure that the 2 read/write heads don't conflict with each other. Not only that, but they would read at completely different speeds due to their different positions on drive itself (the farther out the read head is, the larger amount of data it is able to process in a given timeframe.)