Windows file systems become fragmented over time as file are created and removed. Windows users should therefore run a defragmentation tool periodically to improve disk performance.
Unix file systems, in contrast, do continuous defragmentation, so performance will not degrade significantly over time.
Overwrite performance on some file systems is slower than initial write. Hence, removing files before overwriting them may help program run times.
Most Unix systems offer multiple choices for file systems. Most modern file systems use journaling, in which data that is critical to maintaining file system integrity in the event of a system crash is written to the disk immediately instead of waiting in a memory buffer.
To save time, this data is queued to a special area on the disk known called the journal. Writing to a journal is faster than saving the data in it's final location, since it requires fewer disk head movements.
Journaling reduces write performance, since data is first written to a journal and later moved to its final location. This takes more time and more disk head movements than storing data in a memory buffer until it is written to its final location. However, the performance penalty is marginal if done intelligently. All modern Unix file systems use advanced journaling methods to minimize the performance hit and disk wear.
Popular file systems:
EXT is the most commonly used file system on Linux systems. EXT3 was the first to including journaling, basically as a feature added to EXT2. EXT2 was notorious for incredibly slow file system checks and repairs. The journaling features added by EXT3 greatly reduced the need for repairs, but EXT3 is not the best performer overall and is also hard on disks due to excessive head movements.
EXT4 represents a vast improvement over EXT3 due to major redesign of key components. Performance and reliability are solid.
UFS (Unix File System) evolved from the original Unix system 7 file system and is now used by most BSD systems as well as some commercial systems such as SunOS/Solaris and HP-UX.
FreeBSD's UFS2 includes a unique feature called soft updates, which protects file system integrity in the event of a system crash without using a journal. This allows UFS2 to exhibit better write performance and less disk wear.
XFS is a file system developed by SGI for it's commercial IRIX operating system during the 1990s, which were popular for high-end graphics. SGI IRIX machines were used to develop and featured in the movie Jurassic Park.
XFS has been fully integrated into Linux and is now used as an alternative to EXT4 where high performance and very large partitions are desired.
ZFS is a unique combination of a file system combined with a volume manager, developed by Sun Microsystems.
ZFS is widely regarded as the most advanced file system to date. One of its most unique features is the fact that it does not require partitioning the disk in order to separate file systems. With other file systems, if you want home and /var to be separated and utilize different settings, you must divide the disk into separate partitions. Choosing the optimal size for each partition is almost impossible since we cannot predict the space requirements of the future. With ZFS, you can create multiple file systems, each with its own settings, all of which allocate blocks from the same pool. Thus, you never run out of space in one file system while having unutilized space in others. ZFS also offers advanced software RAID that generally outperforms hardware RAID systems, and many other advanced features such as compression and encryption.
ZFS has been fully integrated into FreeBSD and is now the default file system for high-end FreeBSD servers as well as the GhostBSD desktop system. ZFS does require a lot of RAM, however, so UFS2 is still a better choice for low-end hardware such as net books and embedded FreeBSD systems.