Seagate uses shingled magnetic recording to break capacity barrier, 5TB HDDs coming in 2014
As you probably know, the storage density of a hard drive is utterly astonishing, maxing out at around 700 gigabits per square inch. At that density, each of the tracks (rows or rings) is around 75 nanometers wide, and each of the individual magnetic sites (bits) is is just a few nanometers long. While governed by very different laws of physics, these magnetic sites, just like transistor-based solid-state storage, can only be shrunk so much before they can no longer be read or written accurately. This is where SMR comes in.
In your current hard drive, there is probably between one and three platters that are covered in tracks. Tracks are concentric circles that are broken up into individual magnetic sites that store bits. If the site is magnetised one way (say, north) it is equivalent to binary one; otherwise it’s a binary zero. Now, due to way that magnetism works, there are some very strict rules on a) how big each of these sites has to be to generate measurable magnetism and to be magnetically stable, and b) how close they can be to other magnetic sites, due to interference. Point A was solved by the fairly recent introduction of perpendicular magnetic recording (PMR), and Seagate’s SMR attacks point B.
To prevent interference, there is a gap between each track. This significantly reduces the areal (surface) density, as a large portion of the platter simply isn’t being used. With SMR, the gap is removed, allowing the tracks to overlap each other, like shingles. This massively increases density (25% to begin with, apparently, but more to come), but it introduces some odd issues that need to be rectified with clever firmware.
Basically, your hard drive has a read head and a write head. The read head is very small (so that it can read the magnetism of a tiny magnetic site without interference from other sites/tracks), and the write head is larger. With SMR, because there’s no gap between the tracks, the write head actually writes to the track it’s trying to write to — but also to other nearby tracks that now overlap, as there’s no gap. This is fine if the other tracks happen to be empty, but if data is being stored there, the SMR write process actually destroys the data. To prevent having to re-write the entire disk, tracks are broken up into bands — shortish blocks of tracks — that are rewritten.
This approach obviously isn’t ideal, but Seagate claims that it’s already shipped one million SMR-enabled drives — so the performance hit from having to rewrite data is seemingly minimal. Starting with 2014, Seagate will use SMR to boost platter capacity to 1.25TB, allowing for four-platter 5TB drives. The video above suggests that Seagate can use SMR to reach 20TB, though I’m sure that won’t be for a few years yet. It does make you wonder what happened to heat-assisted magnetic recording (HAMR), though: Seagate has been working on HAMR for more than a decade, reaching an areal density of 1 terabit per square inch last year — but so far, there’s no news of its commercial adoption.
0 comments:
Post a Comment