Ricochet is the best place on the internet to discuss the issues of the day, either through commenting on posts or writing your own for our active and dynamic community in a fully moderated environment. In addition, the Ricochet Audio Network offers over 50 original podcasts with new episodes released every day.
How to Build a Computer 26: Spindles and Platters and So Forth
We’ve discussed what it means to actually store information on a hard disk drive, how you magnetize it and how you pull that information off. Neat stuff, but a bit heavy on the abstract physics. Today we’re going to zoom out a bit and look at the mechanical bits of how hard drives work. Here, let me start you off with a picture. Take a look at these two hard drives (conveniently cracked open for viewing purposes), one I borrowed from the boss man, and the other I picked up off the “Free Stuff” shelf when they moved the engineering department. Tell me which you think stores more data:
None of the above. Neither of is ever going to run again. Look at that dust!
Now I’ve actually got no idea on the history of these two drives, what year they were built, their listed capacity. Couple points of difference tell me the one on the left is older, and stores far less information:
- The platters have a larger diameter
- The brown-orange tint on the disk
- That yellow thing on the side
Right away that color is a dead giveaway; it gets that tint from the recording medium. (And, if I’m being fair, from the reflection of wooden kitchen cabinets. There is a difference in the color, but it’s a bit hard to get on film.) On the older disks, that recording medium was iron oxide. Yeah, rust. Somehow marketing departments weren’t too eager to talk about it that way. “The Corrosion 5,000, with high capacity Rust-based storage!” The copy practically writes itself. The industry moved away from rust because you get better areal density (information per square inch) with different materials, even before that whole magnetoresistance thing.
The yellow thing on the side, by the way, is a ramp. When the hard drive powers down the arms swing over to the side and rest on the ramp. On older hard drives you just rested the thing on the platter. The ramp helps prevent damage from dropping your hard drive, and keeps it from dragging when you’re just spinning your disk up to speed.
Okay, but why the shrunken platters? Don’t we want more information on there? Because we can get so much areal density in other ways; using that extra space becomes less important. “I can store a thousand times more information per square inch, but you only get to use a quarter of the space.” Do you take that deal? Of course; you’re still working with two-hundred and fifty times your previous storage space. But why not go for the full thousand? Larger platters have some drawbacks.
- Speed: Let’s say you’re reading data off your hard drive, and let’s say it’s sorted like my sock drawer (not at all). You grab some from the inner ring, then you go all the way to the outer ring and grab some more, then you’ve got to get back to the inner ring, and so forth. If you have a smaller disk, you’re getting things quicker because you have to wave the arm over less distance. That’s true even if you have a well-ordered sock drawer. You can (if you’re a maniac) take that further. My brother @SamRhody once considered reinstalling all his games to be closer to the hard drive’s spindle to improve speed when gaming. If you’re that worried about it, get a solid state drive (which, to be fair, he did). [Note: Chris B. was kind enough to correct this in comment #6.]
- Flatness: that disk, it isn’t an ideal surface spinning around. There’s always some difference in height. That difference can be really important if you’re expecting your needle to float three nanometers off the surface. It’s easier to make a flat disk when it’s smaller. You have to keep it flat over less area. Smaller disks and shorter arms are also less liable to break from dropping.
- Momentum: It takes a certain amount of energy to spin one of these disks up to the 5400 or 7200 or 10,000 RPM you’re looking for. Pretty obviously that energy is going to be less if you’ve got a smaller platter. Less mass to swing about. And that translates to less energy to keep it spinning, less noise from the motor, and overall a cheaper drive to operate.
On the subject of those motors, let’s talk about ’em. There are two in your drive, one that moves the arm around and one that spins the disk. The one that moves the arm used to be a stepper (a stepper motor is one that allows you to specify exactly how far you want it to spin by changing which coils are active. Lets you do precise controls). When the precision required for positioning got too much for stepper motors the industry moved to voice coils. A voice coil is a fine coil of wire that moves something magnetically. You’ll find them on speakers, and they’re the things that compensate for your shaky hands in your cellphone photos.
(Actually, my company has been building a better solution to that cellphone camera problem. We use something called “shape metal alloy”; a wire that changes shape when you run a current through it. Just putting in a quick plug, and please direct all pricing inquiries to the sales team.)
A suspension under the metaphorical shadow of encroaching doom. And also the actual shadow of poor photography.
The voice coil is the copper oblong ring on the right. If you look back up at the first picture you’ve got a big metal chunk sitting above the voice coil and obscuring it. That’s a magnet. You run current through the copper coils and you generate a magnetic field. That’ll push off of the permanent magnet above it and move your lever. Lever? The circle-thing (in technical terms) right next to it is a pivot; the actual read/write head is way over on the left. You’ll note you get more distance moved out of it than you put into the voice coil. That black thing on the left is a plastic support for the display model, it doesn’t actually show up in a hard drive. Here, lemme give you another photo. This one’s of that older hard drive from the first picture:
Oh yeah, from the side the way the light glints off of the dust motes make’s em real evident. Also the wooden cabinet and the kitchen blinds.
There’s still a black thing, but this one is much smaller. It’s there for its aerodynamic properties. Helps the thing float above the platter on the air currents generated by the spinning disk. You’ll also note that threat are multiple arms for multiple disks. Sort of obvious, in retrospect. Each surface of the disk can contain information (sometimes they leave one blank for marketing purposes). The discussion about why you make the disks smaller also applies to why you don’t just keep stacking disks on there. With one more wrinkle; it’s also harder to keep the various arms in sync the more of ’em you have.
The spindle motor is a servo that spins the disk. It’s got a couple of problems to solve. It’s got to run at high speeds (up to 15,000 RPM in some cases, 7200 RPM is most common), it’s got to maintain that speed very closely, and it’s got to be able to spin up and down to conserve power in between cat videos. (That is, I assume all you degenerates carefully copy all your cat videos to local storage in case YouTube goes down. Some things need to be secured.) The stator is fixed to the disk case, the rotor is fixed to the platters, and in between ’em there’s a ‘fluid bearing’ (oil) to keep the friction down. You can read an interesting bit here about a company extracting data from a hard drive whose spindle has seized up. Sample Quote:
Speaking of lubrication, our data recovery engineers have seen on at least one occasion a previously-opened hard drive whose owner had sprayed WD-40 onto the motor and platters to try and get it spinning up again. It was not a pretty sight.
That about covers it from the mechanical side. I’m figuring on two more posts discussing hard disk drives before we move on to other topics. Join us fortnight next when we discuss recovering data from dead hard disk drives, and how to make sure no one can recover yours, in “Delete my Browser History” or “What, With a Cloth?”
This is part twenty-six of my ongoing series on building a computer, the rough-and-tumble cowpoke way. You may find previous parts under the tag How to Build a Computer. This week’s post has been brought to you by the Singing Cowboy himself, Mr. Gene Autry! Gene Autry comic books are now available in the discount bin at your local comics store. Pick yours up today!
[First – Silicon] [Previous – Magnetoresistance] [Next – Data Recovery]
Published in Science & Technology
Hank, you continue to make inherently fascinating things interesting.
But seriously, great job!
Did you remove the cheese before taking those photos?
I remember the iron oxide color on disk surfaces. I go back far enough that we didn’t even have to take a disk apart to see it. At least that’s the way I remember the 19″ 2MB disks I had to work with.
The bits carved out of cheese and sausage are too small to see in these photos.
Fun fact! Zooming in on your cellphone camera doesn’t work very well for magnifying very small objects.
Ah, something I understand! Uh…actually, something I misunderstood until Hank patiently explained it. Fine post. As usual.
How distressing.
Local copies? Of course, but always remember to keep a quality offsite backup, like Catblaze!
I really appreciate the Backblaze guys for the detailed information that they publish on how to build their custom storage pods (for the intrepid DIY folks) as well as the detailed reliability statistics they generate on consumer hard drives from their massive fleet of drives.
I’m not sure there’s anyone else out there that publishes statistically useful reliability data on hard drives.
Hard drives partition data from outside edge (where the linear velocity of the platter is greatest) inward. Performance falls off as the arm moves inward towards the spindle because each successive circular track is shorter, presenting less area to the heads per revolution of the disk. Enthusiasts would sometimes “short stroke” the disk by setting up partitions to only use the faster outer edges of the platter (the “beginning” of the disk), where you would get maximum data per revolution, and therefore less time wasted on seeks to different tracks.
A fascinating read, Hank. I’m really enjoying your series, and learning something with every article!
Ditto! and BTW, Chris B, you sound like a damn good explainer yourself. You ought to do a post too!
Never heard of that one before. I use Code42, which used to be Crashplan. Didn’t find any comparative reviews of the two just now.
I hope this series never ends, but a Kindle version ought to be in the works. Taken together there’s a helluva lot of information presented in an extraordinarily enjoyable way.
But I think anyone who buys the book should also have Amazon deliver a bunch of salami and cheese.
I spent years as “the backup guy” for an MSP running around 80 backup systems protecting roughly 350 servers. My work has been mostly Windows Server focused, and I have extensive experience with StorageCraft and Datto systems
I have a little less experience with Infrascale, Intronis (now Barracuda MSP), Barracuda, Carbonite, Mozy Pro, iDrive, Backup Exec, Veeam, Windows Server Backup, and Microsoft Azure Backup Server.
I’ve done at least one actual Disaster Recovery with all of those. Never actually used Backblaze, as we aren’t a reseller, but I hear good things about them. I like to read some of their blogs and data publications, and thought the fact that they operate a page themed entirely for backing up cat videos from the web seemed apropos.
That is brilliant. I should have known, making a joke about securing cat pictures, that such a thing must already exist. Hadn’t ever heard of that before now. Thank you sir!
Very well may be. It contradicts what I was reading, but seeing as what I was reading was something published on the internet, it may have not been entirely accurate. (Sorry; I should have warned those of you with weak constitutions to sit down before I made such a stunning statement.)
One of my greatest worries writing this series is that, while I know a little bit about many things, there’s going to be a Ricochetti who knows a great deal about the subject I’m discussing at any given time. On the very first post, for example @GLDIII just happened to know a great deal about how silicon boules (it’s a word) are made.
Much like a newspaper covering something you know about it, once the illusion of competence is shattered I can only hope you’ll forget about it by the next week’s post.
We all have our niche. Your series has taught me at least one thing I didn’t know at all in every article, so if I can enhance your knowledge just a little, well at least I’ve managed to contribute!
I only know this because I had a client with an 2003 SBS Server (that’s the one where Microsoft tells everyone to never put Active Directory, Exchange, and SQL on the same server, then puts all three on the same server). The poor thing just couldn’t keep up with all the traffic hitting it. We couldn’t add RAM because it was 32-bit and had a 4 GB cap.
I eventually did a lot of research and came up with the idea of short stroking the Exchange database. This had a significant improvement in performance. I can’t quantify it, but the server went from “unusable” to just “frustratingly slow.”
When I do Ricochet posts about 3D movies, I note that cameramen (and in the day, they were) used the right eye to peer into the viewfinder while keeping the other eye open to get a peripheral idea what was outside the frame. However, the director of House of Wax, Andre de Toth, was one eyed. How Warner Brothers could find the only one eyed guy in the Director’s Guild to direct 3D I’ll never know. Anyway, de Toth had to use his working eye, the left one, which invalidates that right eye information I’ve been handing out. These things happen.
Yeah, I’ve been there. I’ve also got programs with comments like
“Huh, I thought this would be too slow but it ran fast enough that I didn’t get bored and hit the internet, so that’s a good sign.”
I think I figured it out; that’s not always a good sign. I’ll ask and answer my own question, then you guys can pick apart the answer.
A few weeks ago I posted a picture of the first computer book I ever owned. At that time, long term storage was punched card, punched tape or magnetic tape. Random access memory was made of a fabric of tiny metal rings. In between, there was drum memory, apparently the hard disc of the Fifties. Same idea as a disc drive: continuous rotation, moving read-write heads, special formatting instructions. OK. But disc technology, in the form of audio, was a familiar concept, and the idea of stacking discs wasn’t brand new either.
So why was medium-fast active storage on a drum, even a low inertia hollow one? It stands to reason that if it’s difficult to make a flat disc surface completely orthagonal, it’s gonna be even tougher with a three dimensional object, which is also bound to vibrate more in spinning. Just about everything with a drum looks harder or less precise. So why did we persist with them? I tried as hard as I could to think like Hank Rhody. Like Cliff Robertson in Charly, my IQ rose 50 points as I emulated the sage of Eau Claire. “The root of the square of the cube times the hypotenuse is Planck’s Constant divided by Pi…” and then it came to me.
All the tracks are the same length. No compensation for different diameters. Every track gets its own read/write head.
It’s like so many things in life. You know it’s wrong. But it feels so right.
If it helps, I forget most things within minutes.
Really, the only thing I know about drum memory comes from the story of Mel, the Programmer.
Thanks for providing a real service to the tech-impaired, Hank…Your translation skills are top-shelf/top-notch.
Always faster, cheap, and better, we recall when this use to be the standard:
Remember when you can operate one of these babies you were set for life! -NYTimes Review
We had been using IBM 2420 Mag Tape drives on our IBM 360/50. An upgrade was done, moving to an IBM 370/158. It had 3420’s. One evening, during a particular hectic day, I turned and opened the door of one of them. Uh, oh. It was in a high-speed rewind. We had to call a CE and he was none too happy.
That sounds eerily like something that happened at my university!
It was a surprise to me the drives were not designed for the doors to be locked while they were in operation, especially a high-speed rewind..