Slashdot Mirror
Hitachi Promises 4-TB Hard Drives By 2011
zhang1983 writes "Hitachi says its researchers have successfully shrunken read heads in hard drives to the range of 30-50 nanometers. This will pave the way for quadrupling today's storage limits to 4 terabytes for desktop computers and 1 terabyte on laptops in 2011." Update: 10/15 10:39 GMT by KD : News.com has put up a writeup and a diagram of Hitachi's CPP-GMR head.
Cue the "Nobody needs more than []300GB []1TB []x because I don't have a reason for it" posters
I hope we won't be using hard drives in four years. Let's all pray for a breakthrough in solid-state storage.
It'll get used up fast, with the introduction of Blu-Ray and HDTV movies/tv shows.
It's not really that weird at all. If you have a modest HD movie collection (say, 50), it could easily chew up a big part of the drive. Add a 100gb music collection, maybe half a dozen game installs, OS install, and your 4TB drive suddenly doesn't seem that big.
But that fits a different need - the need for fast access times, low power, etc. This fits its own need - people that need extremely large amounts of storage space, no matter the access time or power usage tradeoffs. Also, while this'll be pretty expensive, keep in mind that SSD drives are still gonna be expensive as hell, and even assuming the price of SSD drives comes down, 500Gb is still gonna cost a pretty penny, while normal mechanical HDD's at that size will probably be no more than $50 dollars (since I can run down to local retail and pick up a 400Gb for about 120 right now).
While its pretty incomprehensible to use even a fraction of the mentioned 4Tb right now, I can see that with high-def video becoming more and more common, at the very least all the people pirating movies and tv shows will use these drives. Also, think about how more and more computers are being sold with TV tuners in them (granted most people will never use them). A few years from now, I can see that instead of regular TV tuners, HDTV capture devices will be much more common - thus people will actually use that space...
Okay, that's great. Hard drives will get bigger. The problem is they aren't getting any faster. I'm having a hard time trying to get RAID 6 working well with my 1TB drives (think rebuild times, RAID 5 will be on its way out). How do I manage a RAID array of 4TB disks that still only give me about 60MB/s real-world write performance. So I put 12 in a RAID 6 and end up with 40TB. How many days will it take to rebuild a failed drive in real-world work loads? Capacity is great - but at some point we are all going to wake up and start begging for faster speeds as well. I think hybrid drives might have a shot, 1TB of flash with 3TB disk might be the right match - but you're still waiting forever on rebuilds (and a policy to manage it).
;-)
I imagine some of you out there, like myself, are starting to see problems with data integrity as the mountain of data you are sitting on climbs in to the petabytes. All I can say is: bit flips suck! Do you KNOW your data is intact? Do you REALLY believe your dozens of 750GB-1TB SATA drives are keeping your data safe? Do you think your RAID card knows what to do if your parity doesn't match on read - does it even CHECK? I hope your backup didn't copy over the silent corruption. I further hope you have the several days it will take to copy your data back over to your super big - super slow - hard drive.
Is anyone thinking optical? Or how about just straight flash? I have a whole stack of 2GB USB flash drives - should I put them in a RAID array?
Its 2007 and we have 1TB drives now... If you apply Moore's law to storage, size should be doubling every 18 months.. that puts 4TB some where around 2.5 years out.
I think 2011 is a pretty conservative estimate.
Whenever I read about advancements in storage space, what comes to mind for me is now there will be NO incentive for companies to ever throw away information they have about you. In years past, physical storage limits--and later data storage limits--has caused companies (and the government) to routinely purge data. With hard drives getting bigger at a rate faster than they can fill them, why expend the effort to get rid of old data? Why would they spend the manhours to delete old data, when it's cheaper just to keep adding larger drives?
The possibly negative consequences here can be very damaging. Imagine the security breach when a company "loses a laptop" that contains 30 years of your transaction history. Or, say you're 20 years old right now, imagine what would happen if in 2040 you decide to run for congress and your opponent pulls out dirt from your Google searches and GMail chats of your youth? Imagine the blackmail material that could be uncovered.
The possibilities are endless, but without a real revolution in the way corporations and government operate, they all seem to lead to the absolute end of privacy.
Is this a news report or a trailer for a motion picture?
Properly disposing data as a business process will take time, but this time will be saved many times over when people don't have to dig up through junk to find what they need, when important things are not buried in crap, when all data worth storing is clean and polished and free of rust,
I'm sorry, but this is just fantasy world 101. I almost never have to look through old mail, but when I do it's because some clients are trying to dredge up something that just not how it happened. Often when I do, it's important that I have all the "useless" mails as well, so you can say with confidence that "No, you just brougth this up two months before the project deadline and it wasn't in any of the workshop summaries [which are in project directories, not mail] before that either."
When I do, it's far more efficient to search up what I need rather than going over old junk - what you're saying is something which would imply that the Internet is useless since it's full of so much redundant, unorganized information. It's quite simply not true, and even though you should extract vital bits to organized systems, keeping the primary source around is very useful.
Extracting experience from current communication to improve business systems (or for that matter, technical routines) should be an ongoing process - it's vital going forward. Going back to old junk to try to figure out what's deletable just to run a "clean ship" is just a big timesink and waste of money. Maybe you'd have an argument if there was a good system not being used because it's all kept as unorganized mailboxes. In my expererience, usually the prolem is there's no such system and doing a clean-up would do nothing to change that.
Live today, because you never know what tomorrow brings
Yes, indeed, we've reached the point where any computer, even if 4 years old, is good enough to do most day-to-day activities (hanging around on the web, wrting some stuff in a word processor, e-mails, and ROFL/LMAOing on AIM/MSN/GMail/Facebook or whatever is the social norm du jour).
Case in point, my current home PC is still Intel Tualatin / 440BX based.
*BUT*...
As you said (and that's something I can confirm here around too), Joe 6 pack buy a new computer every other year, just because his current machine is crawling under viruses and is running too slow (and spitting pop-ups by the dozen). He either pay wads of cash to some repair service that may or may not fix his problems, may or may not lose his data in the process, and he'll have to wait without a machine for a couple of days. Or he gets a new machine. And...
Those outrageous configuration never showed up. Never the less, it seems like Vista was still designed with those in mind.
So in the end the new machine Joe Six pack *WILL* have to be better/faster/stronger, simply because the latest Windows-du-jour has tripled its hardware requirement for no apparant reason.
OS maker will continue to make new versions on a regular basis, mostly because that's their business and they have to keep the cash flow in. Also, there are security issues to fix (by adding additionnal layers of garbage over something that was initially broken by design), legal stuff (add whatever new DRM / Trusted Computing stupidy is latest requirement voted the **AA lobby), add a lot of dubious feature that still 0.1% of the user base will need (built-in tools to sort / upload photos, built-in tool to edit home-made movies, or whatever. Modern OS tend to get confused with distributions and go the Emacs-way of bloat).
All this will result in newer OS that take twice the horsepower to perform the exact same task as older.
And thus, each time Joe 6 pack changes his computer, he gets a newer one, which will obviously have the latest OS on it, and thus will *need* to have 4x the computing power. Just to continue hanging on some IM, sending e-mail, writing things, and browsing porn
"Sufficiently advanced satire is indistinguishable from reality." - [Tips: 1DrYakQDKCQ6y52z6QbnkxHXAocMZJE61o ]
Which CT scanner are we not going to get, in order to pay for all this? Or which MRI will we fail to pay for the last 25% of to pay for all this?
Because, you see, you've just spent your budget on hardware that will never likely be used that gets you no visible day-to-day advantage, except leaving you vulnerable to multiple simultaneous drive failures. (This is surprisingly likely: go read the Google paper on drive failure rates.)
Instead,, you use a second system with snapshot backups, possibly using a system like rsnapshot that supports hard-linked backups. This gives you on-line backup, fast bare-metal restoration, and easy access to yesterday's or last week's data. It also offloads the tape backups. And the mirrored drives can be used for off-loaded backup or mirroring, for creating off-site backup media of actual hard drives, not tapes.
Actually, the scary part is that I can easily see how someone will take it as an invitation to install more bloat on your hard drive, do things even less efficiently, etc.
I started my programming experience almost directly with assembly. Well, I had about a year of BASIC on my parents' ZX-81 first. But that was a damn slow machine (80% or so of the CPU was busy just doing the screen refresh) and Sinclair BASIC was one of the slowest BASICS too. So with that and 1K RAM (you read that right: one kilobyte), you just couldn't do much, you know. So my dad took the Sink-Or-Swim approach and gave me a stack of Intel and Zilog manuals. Anyway, you had to be particularly thrifty on that machine, because your budget of CPU cycles and bytes makes your average wristwatch or fridge nowadays look like a supercomputer.
I say that only to contrast it to the first time I saw a stacktrace (Java, obviously) of an exception in a particularly bloated Cocoon application running in WebSphere. If you printed it, it would run over more than two pages. There were layers upon layers upon layers that the flow had to go through, just to call a method which, here's the best part, didn't even do much. That nested call and all the extra code for reusability sake, and checks, and some reflection thrown in for good measure, obviously took more time than the method code itself needed.
It hurt. Looking at that stacktrace was enough to cause physical pain.
Now I'm not necessarily saying you should throw Cocoon and J2EE away, obviously there are better ways to do that even with them. Like, for a start, make sure your EJB calls are coarse granularity so you don't go back and forth over RMI/IIOP just to check 1 flag.
But how many people do?
The second instance when it caused me pain is when I was testing a particularly bloated XML-based framework, and it took 1.1 seconds on a 2.26 GHz Pentium 4 just for a call to a method that did nothing at all. It just logged the call and returned. That's it. That's 2.5 _billion_ CPU cycles wasted just for a method call. That's more than 30 years worth of Moore's law. Worse yet, someone had used it between methods in the same program, because apparently going through XML layers is so much cooler than plain old method calls. A whole 30 years worth of Moore's Law wasted for the sake of a buzzword. The realization hurt. Literally.
Again, I'm not saying throw XML away generally, though I would say: "bloody use it for what it was meant, not as a buzzword, and not internally between classes in the same program and indeed the same module." It just isn't a replacement for data objects (what Java calls "beans"), nor for a database, nor as just a buzzword to have on the resume.
Each iteration of Moore's Law is taken as yet another invitation to write crappier code, with less skilled monkeys, and don't bother optimizing... or even designing it well in the first place. Why bother? The next generation of CPUs will run it anyway.
And the same applies to RAM and HDD, more or less. I've seen more than one web application which had ballooned to several tens of megabytes (zipped!) by linking every framework in sight. One had 3 different versions of Xerces inside, and some classloader magic, just because it beat sorting out which module needs which version. Better yet, they were mostly just the GUI to an EJB-based application. They didn't actually _do_ more than display the results and accept the input in some forms. Tens of MB just for that.
So now look on your hard drive, especially if you have Vista, and take a wild guess whether those huge executables and DLLs were absolutely needed, or are there mostly because RAM and HDD space are cheap?
At this rate and given 4TB HDDs, how long until you'll install a word processor or spreadsheet off a full HD DVD?
A polar bear is a cartesian bear after a coordinate transform.
...get a real RAID card. Unless you do a lot of tape backups...
I sincerely hope you do backups anyway. RAID is simply there to allow you to continue running a service under some specific failure conditions that would otherwise cause the service to be down whilst hardware is replaced and backups restored - it is not a substitute for backups, RAID and backups accomplish different jobs.
Some examples of failure conditions where RAID won't save you but backups will:
- Some monkey does rm -rf / (or some rogue bit of software buggers the file system).
- The power supply blows up and sends a power spike to all the hard drives in your array (I've personally seen this happen to a business who didn't take backups because they believed RAID did the same job - they lost everything since all the drives in the array blew up).
- The building bursts into flames and guts your server room.
In all these conditions, having a regular off-site backup would save you whereas just using a RAID will not.
http://blog.nexusuk.org