There are a handful of components that every computer has. These components include a processor, memory, motherboard, video card (now optional with an integrated motherboard), and a hard drive. Most everyone pores over their selection of processors, agonizing between the fastest next generation CPU and a cost-effective solution. Video cards run a close second in the pricing game with motherboards and memory given a slight nod to ensure it isn’t last year’s tripe. So, when have you actually looked at the specs on a hard drive? You know, that gizmo that loads your operating system and has all your stuff on it. Yeah, that’s the one. Sure, you’ve got a CAD video card and dual-channel memory, but you do realize that all that data comes from the hard drive, right? If your drive is slow, you’ll be waiting to fill that high speed RAM every single time you load new data.

What about drive types?
Thanks to the wonderful world of well-defined and relatively extensible interfaces like SCSI and IDE/ATA, we no longer have to worry about the drive geometry. To be more precise, I would actually consider it more drive geography or even cartography, because it is a map explaining how the drive stores data and what the controller needs to do to read and write data. However, you really don’t have to worry much about sectors, cylinders, or heads anymore.


You will notice that this article doesn’t touch the most common differentiator of drives: the interface. That’s because if you look for Faithe Wempen’s Daily Drill Down on IDE and my Daily Drill Down on SCSI, you will find an exhaustive breakdown on those topics.

Size and shape
You probably know about the form factor. For hard drives, the typical drive sizes are 2.5”, 3.5”, and 5.25”. Keep in mind this is the size of the platter, rather than the disk itself. A drive is normally about a half inch larger than the platter it contains. The 2.5” drives are targeted for use in laptops, where their small size reduces the power requirements and vibration and improves durability at the cost of limited storage capacity. The 3.5” half-height drive is the current standard containing a number of platters. Full height 5.25” drives are used to increase the number of platters or the size of the platters. However, this larger size comes with a cost due to its significant increase in power consumption and heat generation. On the flip side, it allows an older, less expensive technology to compete with newer drives.

The platters themselves are physically simple in shape and design but complex in manufacture. Basically, they consist of a magnetic film on top of glass, ceramic, or metal platters. They use these materials because they are hard, while 3.5” and 5.25” disks use a floppy plastic substrate. The exact composition of the magnetic film indicates how far the drive heads can be from the platter, how much data you can cram into them (a.k.a. areal density), and how many read/write cycles the drives can survive. The life span of modern drives isn’t as much of an issue as it used to be, with typical mean times before a failure (MTBF) in excess of two years continuous use. However, it can be an issue for servers, but that’s what RAID is for.

The key influence the platter has on a drive purchase is that areal density. Combined with the rotational speed, you get the amount of data that can pass under a drive head. Naturally, that amount varies depending on whether the data is located at the inner section of the drive, where each revolution spins less of the disk under the head, or at the outer edge, spinning the most disk surface possible. This creates a range of data throughputs, which is referred to as the sustained transfer rate.

Transfer rates
So, if you know the sustained transfer rate, then what is the unsustained transfer rate? As you may have noticed, hard drives have a memory cache, usually two to four megabytes worth. Memory, we all know, is much faster than physical drives. So, assuming the data you need is in the drive’s cache, it can be accessed very quickly. During that time the heads are being located onto the correct portion of the platter so that when that cache runs dry, the drives are ready to transfer as much data as possible.

When the data is not in the drive cache, you have to wait for the heads to be aligned. That is what we call latency (or unsustained transfer rate). This increment of time, measured in milliseconds, is the delay your computer will have to suffer through every single time a new request is made to the drive. Oh, sure, the extra 5 milliseconds sure doesn’t seem very long until you realize a 500-Mhz processor can handle 500,000 operations every millisecond. Thus, those 5 milliseconds are 2.5 million lost operations. You can double that calculation for a gigahertz processor. I repeat; for every extra 5 milliseconds of drive latency you will lose between 2.5 and 5 million operations!

Workstation users doing video editing, programming, or CAD can see up to a 5 percent improvement simply by switching to a faster drive. Internet data servers are incredibly sensitive to drive performance where those extra 5 milliseconds per data request can stack up quickly when there are 10 to 20 requests per second on a high-traffic site serving dynamic content.

Drive heads
Of course, it doesn’t matter how fast or large the cache is or how much data you can cram onto the platter if you can’t get it out. This is where the drive heads come into play. Drive heads are small electromagnet devices about twice the size of a toothpick. You know the chattering noise your computer makes? That’s the clicking of the drive head armature whipping back and forth across the disk. In a slow motion world, the heads sweep across the surface of the disk platters like an old phonograph needle. However, if the heads ever touch it, there won’t be sweet music coming out of your drive, but a funeral dirge, as the magnetic film is gouged from the surface. Even if the drive can mark the damaged portion as unusable, those little fragments will eventually get under the heads and cause even more damage. Most drive heads are positioned a mere millimeter or so from the surface, so there is very little leeway available to compensate for mistakes.

Note that a platter has two surfaces, which means two heads. Each surface has one head on top and one on the bottom. This cuts that leeway from a jarred head to just about nil. It’s this very reason that most portable shock-resistant drives used in laptops are so expensive. They have to be built with more rigid drive heads and modified bearings to prevent the platter from deforming under an impact or the head from smacking the platter. In many laptops, the fewest number of platters and drive heads are used to limit the potential risk. Of course, with fewer platters you have, less data and fewer heads means you can’t read as much data at a time.

So now you know
Obviously you have a lot on your mind when it comes to configuring a system. Although processor speed and memory usually take center stage, don’t overlook the importance of the hard drive and its impact on overall system performance. The faster the transfer rate, the more the areal density of the platter, and the higher quality of the drive head all add up to better performance. If you want to rank these factors in order of importance, you should remember that cache is king. Which makes you wonder, how much cache does your cache drive cache if your cache drive could cache data?