The summer slow news silly season is upon us
An article in ScienceNow Daily News reports that Dutch researchers have a lab demonstration of a disk written with polarized laser light. This article was picked up by /. so it will get a lot of undeserved play.
The article says the technology is notable because the bits are written in 40 femtoseconds, 100x faster than a magnetic bit is written. So that would mean that a magnetic bit takes 4 nanoseconds to write, or 250 Mbits or 31.25 Mbytes per second. Sounds about right – until you do the math.
How about a 15k drive?
Seagate’s 3.5″ 15k Cheetah drive is a modern high-performance disk drive.
Google doesn’t know about any independent hardware reviews of the most recent Cheetah. However, reviews of the earlier and lower density 15k.4, find a maximum write data rate of almost 67 MB/s, or over 536 Mbit/s, or more than double the performance postulated in the ScienceNow article. Update: Oops! This is wrong. See update below.
As densities improve with future improvements, bit write times will continue to decline as well. A decade from now, when the scientists hope that a commercial product might debut, disks will have 100x their current areal density, so the claimed 100x advantage in write time will be less than 5x.
How will we get disks to spin 5x faster?
To take advantage of that fast write time, you’ll need to write 5x as many bits in the same amount of time. 75k drives, anyone?
The other problem
/.’s headline writer added another misperception with the headline “100x Faster Hard Drive In Lab”. While this technology, if it ever becomes feasible, would certainly speed up large sequential writes, random reads and writes are dominated by seek times, not write bandwidth. These drives would not feel like 100x or even 5x faster.
Update: Boy, did I screw up! Alert reader Jirko pointed out that 100 x 40 femtoseconds is 4 picoseconds, not nanoseconds. Which puts a different light on the calculations. To wit:
To write 536 Mbits in a second means that each bit takes 1.87 nanoseconds to write. If the minimum time for a magnetic write is 4 picoseconds – and would someone out there who understands this stuff please confirm – then magnetic write times could improve 400x to over 200 Gbits/second (over 25 GB/sec) before the laser time advantage kicks in.
Areal density has been rising about 100x every decade, so all else being equal – which may be a dangerous assumption: are bits square? – linear density should be rising at 10x per decade. In twenty years write times per bit will improve 100x. Since 10x per decade is an annual improvement of roughly 26%, we’re maybe 25 years from reaching the 4 picosecond bit write time posited by the article.
By then HAMR or some descendent will be de rigueur in the fashionable circles of disk drive technology, if we are still using them. End of exculpatory update.
The StorageMojo take
The Dutch research team is to be applauded for their technical feat. The use of polarized laser light is an exciting concept and is, I’d imagine, of interest to researchers in Heat-Assisted Magnetic Recording (HAMR).
Let’s just remember to take a deep breath and think through what such an advance would really mean for data storage. A possible improvement, yes. A breakthrough, no.
Comments welcome of course. And a handy list of the standard prefixes and their numerical value is provided below. I wish I’d had it handy when I started writing.