The external data of a computer are stored file system in the file system in the form of files. The file system is an order and an access system for these data. Access routines for file systems are a component of the operating system. The operating system of the computer becomes likewisein a file system stored.

Files have one file name as well as attributes, which give closer information about the file in a file system nearly always at least. The file names are put down in special files, the listings. Over these listings a file name can andso that a file by the system to be found. A file system forms thus a name area. All files (or file-similar objects) are like that over a clear address (file name inclusive Path or URI) - within the file system - callable. The name of oneFile and further information those the stored data are assigned, than meta data are designated.

For different data media (like magnetic tape, non removable disk, optical data media (CD, DVD,…), Flashspeicher,…) there are special file systems, which consider their characteristics.

Table of contents


historicalthe first paper tapes - (on film later on paper strips) and punch cards - are already files of file systems. They form linear file systems just like tape storages. Later the magnetic drum developed for mass storage and fast access - and fixed disks made optional access possible then for the first timeon arbitrary positions in the file system and thus more complex file systems. These file systems offer the possibility of accessing by names a file. The concept of the file systems was finally so far abstracted that also accesses to files in the net and to devices, thosevirtually as file to be administered, over file systems to be accomplished be able. Thus application programs in a the position are to access these different data sources over a uniform interface.

organization of mass storages

of mass storage devices such as non removable disks -, CD-ROM - and floppy disk drives havenormally a block structure, i.e. from operating system view data can be read or written only as whole blocks. The hardware of the data storage units presents itself in relation to the operating system on a certain level only as large linear surface with many nummerierten blocks.

A blockcovered mostly 512 (2 9) bytes. Modern operating systems combine several blocks for performance and administrative reasons into a cluster of firm size. Today cluster with eight or still more blocks is usually, thus 4096 bytes per cluster. The cluster size is generally a power-of-two number (1024, 2048, 4096 etc.)

a file is a storage surface of arbitrary size, which consists on the equipment of one or more Clustern. In addition each file receives a description structure, those the actual size of, references on the used clusterand possibly. further information such as type of file, owner, rights of access contained knows.

For the allocation from Clustern to files there are thereby several possibilities.

  • The reference of a file consists of the cluster number of the initial cluster and the number on it (physically sequenziell)the following cluster. Disadvantages: when enlargement must if necessary. the whole file to be shifted. This complicates the actual file handling and leads to insufficient performance with many large files. So it can occur that a file cannot be stored, although still sufficientfree memory on the data medium is present.
  • The reference of a file consists of the first cluster number. In each cluster of the file the cluster number of the subsequent cluster is stored. A linked list results. Disadvantages: One does not want the filesequenziell, but z read. B. only the end, must read the operating system in nevertheless the whole file, in order to find the end.
  • Free allocation from Dateiclustern to Folgeclustern by a table on the mass storage (ex.: FAT). Disadvantages: very large description structure,sequenzielles reading or writing somewhat more slowly than ideal, since allocation information is present neither bundled nor with the data.
  • Storage of a field of tuples (Extent Anfangscluster, Extentlänge) in the description structure of the file. A Extent is thereby a consequence of sequential Clustern.Today in many file systems so converted.

Listings contain file names and references to the respective description structures. Since listings are also storage surfaces, particularly marked files are usually used as listings. The first description structure can contain thereby the output listing. Often is first Block for so-called boat block (z. B. Master boat record) reserves, which can be used for the raising of the system. It is not part of the actual file system.

Boat description structures list free cluster cluster with files and listings
example ofthe allocation of a mass storage for a simple file system

access to mass storage

a program accesses the mass storages over the file system. Under UNIX and similar operating systems to it system calls are made available. The most important system calls are here:


  • mkdir - producing a listing
  • chdir - changes into another listing
  • rmdir - deletion of a listing
  • open you - opening a listing
  • READ you - reading from listing entries
  • CLOSE you - latches


  • open, CLOSE - opening and closing oneFile
  • READ, write - creat vintages and
  • letters, unlink - producing and deletion of a file
  • seek - to new positioning write/read pointer

in addition offers the operating system management functions, z. B. for producing a file system on a new data medium, forKonsistenzprüfung, compression or safety device (depending upon operating system and file system differently).

The conversion of the system calls of a program of the Kernel of an operating system are often implemented and to differentiate itself with the different file systems. The Kernel translates the accesses then into the block manipulations of therespective mass storage. (Note: This actually applies only to so-called monolithic Kernel. Modern operating systems however are on a Mikrokernel constructed, so that the file system operations are not implemented by the Kernel.)

if a program a file by means of open open, becomethe file name in the listing searched. The operating system from the appropriate description structures determines the blocks on the mass storage. If a file in the listing is found, one receives also its description structure and thus references to the Clustern and over these to blocks.

With READ the program can access then the cluster of the file (and thus on blocks on the mass storage). If with write the file is increased, if necessary a new cluster is taken out of the free list and in the description structurethe file added. Also the other system calls leave themselves in such a way in cluster and/or. Block accesses translate.

kinds of file systems

linear file systems

those historically first file systems were linear file systems on punched tape or punch card as well as stilltoday for the safety device of data of used magnetic tape systems.

hierarchical file systems

early file systems (CP/M, Apple DOS, Commodore DOS) had only an individual listing, which contained then of references to all files of the mass storage. In most modern file systemsthis listing is the root listing. Here can listings beside normal files also references to further listings, which contain sublists. Also these may have again sublists.

Thus develops a listing structure, which is often represented as a listing tree. The non removable disk drive assembly C:under Windows contain for example beside files like boot.ini and ntldr also listings such as programs, documents and attitudes etc. A listing such as z. B. Then again sublists can contain own files like own pictures or texts. Into texts can thenfor example the normal files Brief1.txt and Brief2.txt stand.

Windows 2000/XP: Mac OS X: Unix/Linux/ZETA: <Drive assembly C:> <Root listing> <Root listing> + - boot.ini | + - [boat] + - ntldr | | + - [Windows] + - [system] | | + - [System32] +[Libary] + - [usr] + - [documents and attitudes] | + - [etc.] | + - [users] + - [home] | + - [user name] | + - [user name] | + - [user name] | + - [own files] | + - [pictures] | + - [pictures] | + - [own pictures] | |+ - Bild1.png | | + - Bild1.png | | + - Bild1.png | | | | | | | + - [texts] | + - [texts] | + - [texts] | | + - Brief1.txt | | + - Brief1.txt | + - Brief1.txt | | + - Brief2.txt| | + - Brief2.txt | + - Brief2.txt | + - <which of bname> | + - <which of bname> + - [programs] + - [Applications] |  <Drive assembly D:> + - [which of bname] listings are marked by [square brackets]. Data media are with <pointed clips> marked.

The listings are called also files and are, depending upon operating system, by backslash \ (DOS, Windows, TOS), Slash / (Unix, Linux, Mac OS X, AmigaOS) or colon :(older Mac OS versions) separately. There itself a hierarchy of listings andOne results in, speaks files here of hierarchical file systems. One calls the way by the file system, indicated by listing names, which are separated with the separators, path. To the file Brief1.txt one knows texts

  • with C:\Dokumente und Einstellungen\benutzername\Eigene files \ \ Brief1.txt (Windows 2000/XP)
  • /Users/benutzername/Texte/Brief1.txt(Mac OS X)
  • access /home/benutzername/Texte/Brief1.txt (Unix/Linux

). With DOS/Windows there are drive assembly letters followed from a colon, which are placed in front the paths within the file system. Each data medium gets its own letter, z. B. usually C: for the first partition thatfirst non removable disk. With Unix there is no drive assembly letters, but only one listing tree. The individual data media are hung up there in certain places in the tree (command mount), so that all data media result in together the total tree. Windows variants, those on Windows NTare based, work internally likewise with such a tree, this tree however opposite the user one hides.

Under AmigaOS a mixture of the layers takes place of DOS and Unix. The drive assemblies designated after Unix nomenclature are addressed with colon (fd0: , hda1: , sda2:). In addition itcan “logical” colon drive assembly designations such as LIBS: by ASSIGNS independently of the “physical data medium” to be assigned.

Frequently the term file system does not only designate the structure and the kind, as the data on a data medium are organized, but generally the whole tree with severaldifferent file systems (non removable disk, CD-ROM,…). One would have to speak correct way here of a name area, which by different partial name areas (merged data media with their file systems) is formed, since however this name area is very file-referred, frequently only of the file system is spoken.

Work on []

Network file systems

the system calls such as open, READ, etc. can be transferred also over a network to a server. This accomplishes then the accesses to its mass storages and returns the delivery the requested information at the Client .

There the sameSystem calls to be used, differ the accesses from program and user view not from on the local devices. One speaks here of transparent accesses, because the user does not see the detour on the other computer, but apparently directly on thosePlate of the distant computer looks - as by a transparent windowpane. For network file systems special network minutes are available .

Can on a file system e.g. in a SAN of several systems becomes, speaks one accessed parallel directly of a global one - or Cluster file system. Additional measures are to be seized, in order to avoid data corruption by mutual overwriting. In addition a meta data server is used. The meta data accesses - typically over a LAN - pass all system on to the meta data server, these operations such as listing accessesand block and/or. Cluster assignments makes. The actual data access been made then by the SAN as if the file system locally would be attached. Since the Overhead falls in particular by the transmission to the meta data server with large files hardly important, so one can Transmission rate similarly the one directly attached file system to be realized.

special one virtual file systems

the open - READ - model can be applied also to devices and objects, which are not addressed normally via file systems. Thus that becomes access upthese objects identically to the access to normal files, which brings usually advantages.

Under the present Linux Kernels (and. A. ) System and process information about the virtual can version 2.6 be queried and changed proc - file system. The virtual file /proc/cpuinfo e.g. supplies.Information about the processor.

Proc there are such dummy file systems as under Linux some: In addition count sysfs, usbdevfs or devpts; under some BSDs it gives kernfs. Not least devfs and/or. udev for equipment files (see below). All theseFile systems contain only purely virtually existing files with information or devices, which is illustrated to a “file”.

The Kernel gaukelt here quasi the existence of a file forwards, as she could be present on a mass storage also.

Files in ramfs or tmpfs, in addition, initrd and to similar file systems however actually exist, however only in the memory are held. They are used for speed reasons and for logical-technical reasons during the boat phase.


many modern file systems have the principle thatFile generalizes, so that one in a file not only a consequence of bytes, a so-called Stream (English. River), but several of such consequences (Alternate DATA Streams) to store can. It is possible to work on parts of a file withoutpossibly existing other parts, which can be very large, shift to have.

The support lacking of multiple Streams is problematic. That expresses itself on the one hand in the fact that alternative data with the transfer onto other file systems (ISO 9660, FAT, ext2) without warninglost it goes, on the other hand into the fact that hardly a tool supports these, why one cannot see the data stored there so easily and for example of virus scanner viruses stored there surveyed.

Thus that the main data stream of changes at the other rivers notone affects, arise advantages for the performance, the space requirement and data security.

Under Inode - based file systems are left possible savings SE files , hardon the left of and symbolic. Also technically differently developed file systems know recently partially these characteristics.

File systems from the Unix rangeknow special equipment files. Their name is often fixed by conventions thereby, it can as desired be renamed; so have e.g. also the keyboard, mouse and other interfaces special file names, on which with open, READ write , is accessedcan, even main storage has a file name (/dev/mem). (Unix philosophy in addition reads: „Everything is a file, and if not, it should be a file. “)

In other file systems (as under MS-DOS the FAT - systems were based, in addition, under NTFS)it gives firmly programmed “magic” file names with special meanings, which are similar to the system of the equipment files. These special file names were in the past repeatedly cause of safety problems, there the appropriate names partially unite applications do not admit were and thereforewere not filtered, but partially also because the access protection was insufficiently regulated on the devices associated with it.

data base file systems

new concepts for file management are data base-based file systems. Instead of in a hierarchically developed administration, files become on the basis their characteristics,like type of file, topic, author or similar Meta - information identifies. The formulation of a file search can take place therefore in SQL or in natural speech.

Beginnings for it are GNOMES STORAGE, WinFS and BFS. BFS is developed already finished and offersappropriate features of a data base. In systems the BeOS and ZETA is used as standard file system.

Such a form of storage on the basis Meta - information depends always on the co-operation of the users, who supply the system with suitable Meta information.

examples of file systems

many early operating systems (z. B. CP/M, Apple DOS, Commodore DOS) had only in each case one file system, which did not carry an own name. These are to be marked if necessary with the manufacturer more exactly.

More modern file systems:


  • ADFS (Acorn StrongARM)
  • AdvFS (HP Tru64) cluster file system with journaling
  • FFS Berkeley nearly file system - variant of UF under BSD
  • CFS Crypto file system - on a NFS Daemon putting on coded file system
  • EAFS Extended AcerFile system - use under SCO Unix
  • EFS (Extent Filing system predecessor of XFS under IRIX)
  • EncFS Encrypted file system - on FUSE based user country file system
  • ext2 (second extended file system, long time the Linux - standard file system)
  • ext3 (developed furtherVariant of ext2 with journaling)
  • FUSE user country - file system in the Linux Kernel
  • of GNOMES STORAGE (data base-based file system)
  • HTFS High throughput file system - standard file system under SCO open server
  • JFFS2 (Journaling Flash file system, version 2) Flash memory file system for Embedded Linux employment
  • JFS (JournaledFile system of IBM)
  • LUFS Linux user country file system - hybrid user country file system
  • NFS Network File System (developed by Sun for Solaris)
  • NILFS (a quite new file system of the NTT with journaling)
  • minix (of the operating system of the same name)
  • OCFS2 Oracle cluster fileSystem for Linux Version2 - POSIX compatible cluster Dateisystm
  • ReiserFS (Linux Journaling file system of Hans Reiser)
  • s5fs (the classical file system system of the V - Unix of AT&T)
  • Files-11 (OpenVMS Files-11 file system, standard file system under OpenVMS)
  • UF (UNIX file system, uses under Solaris and BSD)
  • VxFS (Journaled file system of VERITAS)
  • xFS network file system
  • XFS (a further journaling file system of SGI)
  • ZFS 128-Bit-Dateisystem starting from Solaris 10
  • SQUASH compressed READ Only file system


  • FAT and/or. FAT12(File Allocation Table, for disks)
  • FAT16 (extended FAT system for non removable disks)
  • FAT32 (extended FAT for large non removable disks)

Ms-Windows (starting from Windows 2000) supports all MS-DOS - file systems, additionally:

  • VFAT (Virtual FAT, longer file names support, for all FAT systems - support starting from Windows 95)
  • NTFS (Journaling Dateisystem - support starting from Windows NT)
  • WinFS (for the future announced data base-based file system)


  • OFS (Amiga old file system)
  • FFS (Amiga nearly file system)
  • SFS (Smart file system) standard under MorphOS



  • JFS (Journaled file system)
  • HPFS (High performance file system)


  • BFS
  • OpenBFS (64 bits, multithreaded, journaliertes, data base-similar file system)


  • ISO9660 (CDFS - file system for CD-ROM,after the slogan „smallest common denominator “)
  • Joliet (extension of the ISO9660 of the company Microsoft)
  • Rockridge (extension of the ISO9660 for UNIX)
  • UDF (universal disk format, and. A. on DVDs of all types common)


  • AFS (Andrew file system)
  • Apple Filing Protocol (over networks attached file system particularly for Mac of OS systems)
  • Coda (a advanced network file system similarly to NFS)
  • DFS (distributed file system the open Group, an advancement Andrew of the file system; the same abbreviation designates the system of the same name of Microsoft)
  • NCP (NetWareCore Protocol)
  • NFS (Network File System; a file system particularly for Unix - well-behaved systems) SMB/CIFS (
  • a file system particularly for Windows -, attached over networks, systems, attached over networks)
  • xFS (a distributed and decentralized network - file system)

other one:

  • WAFL (Write Anywhere fileLayout) of prop. guessing acre file system on the NAS - systems of the company Network Appliance. Structurally partly with FFS and/or. IBM TransArc episode file system comparably.

With some the file systems specified above concerns it Journaling Dateisysteme. All file systems have together that upit also of other vendors' systems to be accessed can, if the operating system this is supported directly or it is made possible for the operating system over appropriate driver often commodity. Exceptions form file systems, which support an extended authorization, the possibility of the coding offer, or for their exact function mode in Trade secret is (for example NTFS).

Web on the left of


  > German to English > (Machine translated into English)