Why Does Partition Placement Matter

Hard disks store data on circular disk platters. Each platter contains a large number of tracks, which are in turn broken up into sectors. The disk platters rotate at a high velocity (typically 5,400 to 10,000 revolutions per minute, or rpm). A pivoting arm carries a head that reads and writes data from individual sectors as they pass beneath it. Most modern disks have two or three platters stacked atop one another. The tracks in these drives are logically grouped into one mass that's called a cylinder. The read/write heads for the drive move in a synchronized fashion so that all the tracks that make up a cylinder are accessed at once. Outer tracks are broken into more sectors than are inner tracks, in order to make better use of the greater area in outer tracks. This arrangement is illustrated in Figure 2.1.

Figure 2.1: Hard disks are built from platters, each of which is broken into tracks, which are broken into sectors.

Note In order to retrieve data from a disk, the disk must be told which cylinder, head, and sector to read. This is the basis of the so-called cylinder/head/sector (CHS)

Figure 2.1: Hard disks are built from platters, each of which is broken into tracks, which are broken into sectors.

Note In order to retrieve data from a disk, the disk must be told which cylinder, head, and sector to read. This is the basis of the so-called cylinder/head/sector (CHS)

geometry that IA-32 computers have traditionally used. Modern drives create a fictitious CHS geometry to satisfy IA-32 BlOSes, because standard IA-32 BlOSes aren't set up to handle the variable number of sectors per track used by modern drives. Modern drives can also use a linear (or logical) block addressing (LBA) mode in which sectors receive sequential numbers and the drive determines the correct cylinder, head, and sector. LBA mode is slowly taking over from CHS mode, but CHS mode must still be supported to keep everything working together.

You can think of partitions as occupying concentric rings of a hard disk. In order to access data from one partition or another, or indeed from different areas within a single partition, the disk head must move to read the appropriate cylinder. This movement takes time, which reduces disk throughput immediately after such an operation, which is known as a seek. Seeks take differing amounts of time depending upon the distance that must be traversed; moving from the innermost cylinder to the outermost cylinder takes more time than seeking just halfway, for instance.

The result is that, for optimal performance, you should place partitions so that most data accesses are within as narrow a range of cylinders as possible. This practice will minimize seek times, thereby improving performance. Another important feature is that the data density is greater on outer cylinders than on inner cylinders, which results in better disk performance on outer cylinders. The outer cylinders are the early ones, from the point of view of the operating system. You can use this fact to your advantage when placing partitions.

0 0

Post a comment