Size, seek time, rotation speed, mean time between failureJun 06 '00 (Updated Jun 08 '00) Write an essay on this topic.IDE hard drives are currently sold based on several parameters: size (in gigabytes), seek time (in milliseconds), rotation speed (5400, 7200, or 10000 rpm), buffer size (512k, 1 meg, 2 meg) and specific interface type (ATA/33, ATA/66). Mean time between failure is also an important parameter, but hard to find information about. SIZE If you started typing 50 words per minute at birth and continued 24 hours per day until you died, you would not fill any currently available hard drive. By any reasonable standard they are huge. The things that will fill your hard drive are large pieces of software (especially games), digital video (huge), digital audio (mp3 is roughly 1 megabyte per minute = 20 hours / gigabyte), and high resolution digital pictures (roughly 1 megabyte each for a 3.3 megapixel JPEG). You can therefore estimate your own needs. In my case: Windows = 1 gig, Microsoft Office and other apps = 1 gig, a few games = 3 gigs, 20 hours of music = 1 gig, and 1000 pictures = 1 gig for a total of 7 gigabytes. When you buy storage, estimate the amount you need, then DOUBLE it. Yes, double it. So I actually bought a 15 gig drive. SPEED Few people talk about the speed of a hard drive. Speed is determined by two factors, latency and bandwidth. Latency is the time between when the computer requests a chunk of information and the first bit of it appears. For hard drives, this is determined by two things: how long it takes to move the read head over to the track you need to read, and how long it takes the disk to spin around until the data you want is under the read head. The sum of these two numbers is the "seek time." Bandwidth is how fast the information flows off the drive once it gets started. This is limited both by the type of interface (ATA/33 = 33 megs/sec, ATA/66 = 66 megs/sec, ATA/100 = you get the idea) and how fast the disk spins. I don't know if current drives spin fast enough to completely saturate these levels of bandwidth, but I suspect so. So what's important? Listen to Windows as it starts up. Hear all that thrashing around? That's the head moving back and forth on the disk. If you want to make this kind of access quicker, go for low seek times, which probably will involve higher RPM speeds. Now duplicate one giant file on your disk. The disk doesn't thrash, but the process takes time. This is limited by disk bandwidth. To speed this up, buy BOTH a motherboard and a disk that support ATA/66. For most people, latency is much more important than bandwidth. The exception is digital video -- sometimes this requires high bandwidth disk to keep up with the video. So I recommend paying money for low seek times. Small differences in seek time make a big difference. 6 ms and 9 ms don't sound that different, but 9 ms is 150% of 6 ms. Doing things that require constant seeking (like starting the computer) can take up to 50% longer on the 9 ms drive. Rotation speed is the major component of seek time. Let's say you want to read some data that the read head just passed over. You have to wait for the entire disk to rotate once to read it. Assume 6000 rpm = 100 rotations per second = 10 ms per rotation. That's a 10 ms wait worst case, and a 5 ms wait on average. That's why you don't often see seek times below 5 ms unless it's a 10,000 rpm drive. BUFFER Hard disks remember the last 1 meg or so of data you read. If you ask for it again, the disk gives it back instantly. This sounds useful, but Windows duplicates this process in the "disk cache," which is in main memory and is much much faster than talking to the disk. Plus, the chance that you'll ask for something that's in the buffer doesn't increase all that much as the buffer goes from 512k to 1 meg to 2 meg. I wouldn't pay extra for a larger buffer. MTBF = Mean Time Between Failure All hard drives fail. Yes, even the disk you are using right now. Several things determine this. One is use -- if you have a server that is pounding its disk all day, it doesn't last as long as one that spends most of its time spinning idle. Unfortunately, "use" also includes the number of times the disk is spun up and spun down. I don't know how many manufacturers report this sort of data. I suspect MTBFs are in the high tens of thousands of hours (10,000 hours = 1 year). So you might get a decade, on average, before your disk dies. Hopefully you'll buy another bigger, cheaper disk every five years and copy the information over. NOISE and HEAT Hard drives make noise. Hard drives also produce heat. You won't really know if either of these will be an issue until you buy the drive. Just pay attention once you get it -- a loud hard drive can make you crazy. I hope this helps you interpret data you read about hard disks.  |
| Read all comments (1)|Write your own comment |
|
Ads by Google
|
