Defragmentation - User and Performance Issues

User and Performance Issues

In a wide range of modern multi-user operating systems, an ordinary user cannot defragment the system disks since superuser (or "Administrator") access is required to move system files. Additionally, file systems such as NTFS are designed to decrease the likelihood of fragmentation. Improvements in modern hard drives such as RAM cache, faster platter rotation speed, command queuing (SCSI TCQ/SATA NCQ), and greater data density reduce the negative impact of fragmentation on system performance to some degree, though increases in commonly used data quantities offset those benefits. However, modern systems profit enormously from the huge disk capacities currently available, since partially filled disks fragment much less than full disks, and on a high-capacity HDD, the same partition occupies a smaller range of cylinders, resulting in faster seeks. However, the average access time can never be lower than a half rotation of the platters, and platter rotation (measured in rpm) is the speed characteristic of HDDs which has experienced the slowest growth over the decades (compared to data transfer rate and seek time), so minimizing the number of seeks remains beneficial in most storage-heavy applications. Defragmentation is just that: ensuring that there is at most one seek per file, counting only the seeks to non-adjacent tracks.

When reading data from a conventional electromechanical hard disk drive, the disk controller must first position the head, relatively slowly, to the track where a given fragment resides, and then wait while the disk platter rotates until the fragment reaches the head.

Since disks based on flash memory have no moving parts, random access of a fragment does not suffer this delay, making defragmentation to optimize access speed unnecessary. Furthermore, since flash memory can be written to only a limited number of times before it fails, defragmentation is actually detrimental.