Super User is a question and answer site for computer enthusiasts and power users. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I want a partition for storing temporary files (such as pagefile.sys on Windows, or linux swap partition), and I want this partition be faster than others.

We all know that the platter spin speed of the outer side of the platter is faster than the inner side. Almost all partition software use a bar chart to show partitions, thus I don't know how partitions are physically located on HDD platters. As a result I don't know which partition is faster.

Is the physical shape of a partition on the platter like a ring, or a piece of pie (as illustrator on this site )?

What about multiple platters cases? For example: paritioning a 2TB HDD which has 3 platters into 3 partitions with same size. Does each partition occupy a single platter, or does each partition consist of several rings/pies from each platter?

share|improve this question
up vote 3 down vote accepted

The purpose of having multiple platters is not only to go beyond the limitations of storage density, but also to increase throughput. The heads on a hard drive move together across all the platters (even on a single platter drive, there are two heads moving together on the top and bottom). This means when the heads are on a particular cylinder (the term used to refer to the tracks aligned at the same location across multiple sides and platters), the information from multiple platters can be read or written one after the other, just by electronically switching from one head to the next, without physically moving the heads to another track (although micro-correction in head position is often necessary).

The performance of magnetic hard drives depend mainly on three factors:

  • Rotational latency (depends on the speed of rotation, like 5400rpm, 7200rpm, 10000rpm and so on)
  • Seek time (the average duration taken by the heads to move to a particular track)
  • Buffer memory (the higher the buffer, the better)

Considering the above factors, it makes sense that data is written across platters on a single cylinder first before moving on to the next cylinder so that the impact of seek time is minimized. Expanding this to partitions, the allocation should be ring shaped (covering several cylinders across platters) to maximize performance. This is how the addressing mechanism (LBA) works on hard drives.

The pie shaped schematic shown in the site you mentioned is not how it is allocated. Such an allocation would cause much poorer performance.

share|improve this answer
I guess the LBA and CHS equivalence table in LBA CHS conversion section also shows the same result. When LBA address increasing, sector will increase first, then head, then cylinder. – LiuYan 刘研 Oct 24 '12 at 16:57
Yes, that's why the explanation given in the answer is appropriate for the question. – M K Oct 24 '12 at 17:37
> the information from multiple platters can be read simultaneously and processed by the drive controller. No, that doesn't happen. Regardless of the number of platters, there is just one set of read and write circuits. It is switched to the head for whatever surface is being accessed. So just one surface is being read or written at a time. – Jamie Hanrahan Oct 8 '14 at 22:31

I want a partition for storing temporary files (such as pagefile.sys on Windows, or linux swap partition), and I want this partition be faster than others.

That isn't really the right way to think about the pagefile (or swapfile). These files are used by the OS for infrequently-accessed stuff. You shouldn't be hitting your pagefile often enough for tweaks like this to matter. If you are, the right answer is to add more RAM, not worry about "what is the fastest partition".

By the way, since the advent of Zone Bit Recording, the data transfer speed for the outer cylinders is indeed faster. But the actual data transfer is almost always the shortest part of an I/O operation, so this difference doesn't affect total throughput much. But also, as a side effect of ZBR, there is more data in any given outer cylinder than in an inner. So using outer cylinders will reduce the amount of head movement you have to do to access a given amount of data. This is a much bigger win.

share|improve this answer

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .