Lament for the hard drive

TL;DR: We don’t have petabyte drives because who the fuck needs a petabyte drive?

4 Likes

well outside of a data center not so much and i am willing to be we will have petabyte solid state drives before spinning disks.

3 Likes

I wouldn’t mind one.

Cloud sucks and reliance on third parties for streaming sucks even more. One gets dependent on capricious third parties relying on capricious market to stay in existence.

We need at least 20-40 TB disks, reliable and inexpensive, and we need them now.

15 Likes

True, the trend has had various periods of growth; inductive heads, vertical recording, Viterbi encoding, Reed Solomon error correction, MR heads, GMR heads; and things have slowed down recently, but the overall trend compares well with Moore’s law, and the industry has constantly been pushing the boundaries of math, physics and technology. Solid state technologies like bubble memory have been threatening to replace rotating media since I was designing read/write channels in the 1980s. They’ve come a long way (I like my little USB3 500GB SSD) but have still not won the price/capacity/performance race.

https://www.ibm.com/developerworks/community/blogs/storagevirtualization/entry/a_brief_history_of_access_density1?lang=en

3 Likes

http://imgur.com/LrrFLFC

3 Likes

I went on a really interesting tour from one of the engineers at Seagate of their research and development facility. One of the things she said was that while home use of hard drives is on the decline, the total number of hard drives allocated to your life is probably higher than ever due to the massive number of drives in data centers. They are selling more 3.5" factor drives than ever before, just not to home consumers.

The other big point she made is that higher storage density is a low research priority right now because we just don’t need much higher density in the data center market - they need more reliability and lower energy usage. Instead of touting the highest capacity drives when selling to data centers, they now hype their longer MTBFs and lower GBs/Watt, which are the desired features for the new market.

18 Likes

There’s that. I agree with you on that.

But have you considered that many, many optimistic engineers believe in
that Singularity foolishness?
That means developing better connectivity and such, I would think.

1 Like

One man’s foolishness is another man’s inevitability.

Remember, airplanes heavier than air cannot fly. And rocks cannot fall from the sky because there are no rocks up there.

2 Likes

Once SSD first got to a sane price in … geez around 2009, regular hard drives became ridiculously slow in comparison. It is the difference between a modem and a broadband internet connection. You can’t go back to that kind of brutally slow speed. You can’t.

Now some higher end consumer ssds are reaching over 600 megabytes/sec of read and write. More than 1tb/sec bandwidth is not far off either. I would absolutely rather have 256gb of SSD storage that is crazy fast, than 4 TB of old school spinning rust storage that is two or three orders of magnitude slower.

So yeah at some reasonable level, screw size. I will take speed, and any other sane computer user would, too.

9 Likes

That is precisely why I still play vinyl. It makes you a minor but active participant.

4 Likes

The small act of commitment to a period of passivity is why I still go to the cinema, mutatis mutandis: I’ve made a decision to sit and watch what’s going to be shown me for the next couple of hours.

2 Likes

Well that and the technology is now limited by the media the spinning media spins in… and the drive filled with helium suck. 10TB drives are $14K and are basically held together with gum wrappers.

Since the first small (20GB or so) SSDs came out, every computer I’ve had has run its OS on one, whether pre-installed or retro-fitted. Spinning rust is for /data and such, and if I had the money I’d be rid of it entirely.

2 Likes

We use rifles. Seems to work.

5 Likes

Well, if you’re talking about end of life for the drive, kinetic is probably the best way to go.

5 Likes

SSD failure is something I get quite a bit more frequently than platter failure so I keep my data on spinning platters (plus Spideroak). But yeah. Wouldn’t dream of running an OS on spinning platters unless it was a mission critical, must only rarely fail, “speed be damned” situation.

But I agree with everyone arguing for large amounts of local storage as contradictory as that might sound. Until U.S. ISPs are dragged kicking and screaming into modern speeds and infrastructure (and, let’s face it, U.S. IP law allows me to stop paying for all my movies and TV shows over and over again), local storage has an important use.

3 Likes

Really?

2 Likes

A life of spinning rust backed data ain’t a life worth living any more, man.

2 Likes

can’t we have both? :slight_smile:

I have 2 SSD for frequently used data, like apps, os, and current projects…and a RAID for storage of less frequently used data and archives.

6 Likes

Correct. And speed.
I work in IT infrastructure and the changes have been fun to watch over the last decade. Of course, virtualization is the standard now, which means shared storage on a dedicated array or via hyper-converged like Nutanix.
We’re looking into some new arrays sold by Kaminario which are FC and ISCSI SSD’s. They are extremely impressive and in a large environment, now comparable in price to the other guys selling hybrid arrays.
If you’ve got a big Oracle or SQL backend running on VMware, that kind of thing kills. Same with a large VDI deployment.

6 Likes