SSD vs. Rotating Media

Well if HDDs grew faster than Moore's law then the 60 MB drive of 1989 would be 60 TB by now.

So we are pretty close to that, no? I have a 5 TB drive sitting on my desk. Or do you mean they would have to be bigger than 60 TB?

I looked this up and earlier disk drives were doubling in capacity every

Especially since they'd take a month to back up!

Cheers

Phil Hobbs

Jon

There are some "interesting" (in several senses of the word) issues that come up when you try to push hard-drive capacity much higher.

A lot of the problem results from the fact that at today's capacities, the physical size of each bit storage area (each individual magnetic domain) is *tiny*. The domains are so small that they're beginning to suffer from various magnetic effects, such as:

- thermal self-erasure. The amount of energy stored in each bit domain's magnetic field is so small that random (thermal) fluctuations can "flip" it, thus changing the stored bit.

- adjacent-track erasure. The bits are so small that it's very difficult to construct a read/write head whose electromagnetic field doesn't "fringe out" and affect adjacent tracks when writing. With today's drives, writing data to a sector actually weakens the field of bits stored in the adjacent track, and over time this causes the stored data to become harder to read. The drives have to spend a fair bit of their time "remembering" how frequently each track has been weakened by writes to its neighbor, and periodically reading the data back, moving it to a safe location, and then rewriting it "in place" to restore the data strength.

Look up "shingle recording" to see where things are going, as a result of this.

- physical fragility. In order to write smaller bits (closer together in both X and Y dimensions) and keep self-erasure down to just a nuisance, you have to have a smaller read/write head, and "fly" it closer to the platter. With less room between head and platter, the mechanical tolerances have to be finer and the head becomes more vulnerable to interference from "crud" (particles of dust, smoke, or debris from within the drive) on the platter.

And, yes, as you say, there isn't a huge demand in the consumer space for really huge drives. The ultra-high-capacity drives tend to be sold into the "cloud storage" market, moderate-sized-but-wicked- fast drives go into function-specific servers, and "operating system and personal storage" drives for home and laptop use that don't need such high capacities are becoming predominantly SSD rather than hard drive due to the speed and robustness advantages.

I back up hard drives to other hard drives.

A TV-STB for recording all the TV-programs that you never have time to watch :-)

Yes, I'm doing this for daily backups, but I want some off-site and in the fire safe backups, too. So far, what I need fits on a couple plain DVD's, but it is getting tight. I'm looking at getting a blu-ray writer.

Jon

That's a function of the data you wish to back up, not the drive capacity. If you want to back up the entire drive the only really practical way is to another drive. That reminds me, I need to do a back up.

If hard drives were still the same physical size as in 1989, 60TB might not be a problem, aside from the insane cost of such a large drive.

A few days ago I came across a box of full height hard disks in sizes from

Customer got some SSD raid mirrored disks, against my advice. After losing a couple of them, they switched back to magnetic disk. I told them that they could pay for the most expensive SSD, but the chance of having a birthday is still 50/50.

I didn't know they made full height hard drives at a time they could make a 9 GB drive. Maybe you mean 3.5 inch full height drives. The old

We do offsite backups onto terabyte USB drives. Once a month, we dump everything onto a drive, put it in a ziploc bag, and stash it in one of four offsite locations. They are used write-once, assuming that we'd ever only read one if we have a disaster to recover from. A terabyte USB drive is now about $55.

We have other backups for shorter-term stuff, nightly and weekly.

It's more that the increase in capacity was driven by the discovery of giant magneto-resistance

I knew one of the guys that worked in the area - Theo Rasing - who got a (Dutch) Spinoza Prize. Nice guy.

Colossal magneto-resistance is something different, and doesn't seem to have been exploited yet

Spinning rust mills have very high latency. No one wants them. Intel/Micron X-Point will also be available soon.

I do! And all the server farms do, too.

There is a long and tragic history of phase-change memory. This will be interesting.

Depends on what you're using them for. Their three shining virtues are (a) they don't wear out from read/write cycles; (b) they're much easier to erase reliably; and (c) they're much cheaper per GB.

Advantage (b) is related to the wear-levelling algorithms that flash needs to reduce its tendency to wear out.

I expect to be using HDDs for awhile yet. (BTW they haven't used iron oxide for a good 20 years.)

Cheers

Phil Hobbs

On Wed, 27 Apr 2016 19:58:39 -0700 (PDT), snipped-for-privacy@gmail.com Gave us:

"Spinning rust mills"? You're an idiot. Aluminum case, Aluminum platters, Stainless steel head arms. What's to rust, goofy child?

Magnetic disc "hard drives" are going to be around for quite a while yet.