Large files (large). If the data file is placed in one account MET, that fact is
reflected in the title attribute Data, which contains a sign that this attribute
is a resident, which is located in segments outside tables MFT. In this case,
Data attribute contains the address information (LCN, VCN, k) each segment data
(Figure 7.21).
Fig. 7.22. Huge file Extreme files (extremely huge). For extreme file attribute
Attribute List can specify multiple attributes of the additional entries MFT (Figure
7.23). You can also use a double indirect addressing, when nerezidentny
attribute will link to other non-resident attributes, so NTFS attributes may not
be too much for the length. Very large files (huge). If the file is so great
that its attribute data store addresses resident segments data is not stored in
one record, this attribute is placed in another account MFT, a reference to
attribute this to the main entry file (Figure 7.22). This link is contained in
the attribute Attribute List. Sam attribute data continues to address resident
data segments.
Catalogs NTFS Each directory NTFS is one entrance to the MFT table, which
contains an attribute Index Root. The index contains a list of files in the
directory. Indices can sort files to speed up the search based on the value of
certain attributes. Usually, file systems, files sorted by name. NTFS allows for
sorting any attribute, if it is stored in a resident. There are two forms of
storing the list of files. Small directories (small indexes). If the number of
files in the directory is not a list of the files can be resident in writing MFT,
a directory (see Figure 7.24). For a list of resident storage is the only
attribute-Index Root. File contains a list of file attributes. By default, the
filename, as well as the record number MTF, a record primary file.
Large catalogues (large indexes). As the directory grows, the list of files may
require resident of storage. However, the initial list is always resident in the
root directory entry in the table MFT (Figure 7.25). Filenames resident of the
list of files are called nodes Finally tree (binary tree). The rest of the list
of files located outside MFT. Their search is a special attribute Allocation
Index, representing addresses segments, the remaining holders of the list folder.
Some of the lists are the leaves of a tree, while others are intermediate nodes,
which is contained along with the names of files attribute Index Allocation
indicating lists of files at lower levels. Sites binary tree divided the entire
list of files into several groups. Debug output log file name of each node is
the name of the last file in the group. It is believed that the file names
compared leksikograficheski, it is first taken into account codes first two
characters compared names, the name is less if the code of the first character
is less arithmetic value, equality code first characters compares codes second
character names and e. For example, the file f 1. ehe, the first node of the
binary tree shown in Fig. 7.25 is the name leksikograficheski big names avia.exe,
az.exe ,..., emax.exe forming the first list of names directory. Accordingly
file ltr.exe has addressed most of the second group of names, and all files with
the names, big ltr.exe, a third and final group. Search Directory unique
filename, which is NTFS number of major record in the MFT file, it is a symbolic
name to read. First, a symbolic name searches compared to the first node in the
name of the resident part of the index. If fewer original name, it means that it
has to be sought in the first resident group, which of the attributes Index
Allocation pulled to the segment (VCNj, LCN ^ K!) , Keeps files first. Among the
names of the group is done directly display the names and compared to a
convergence of characters from the original name stored in the directory name.
If convergence is extracted from the catalogue number of major record file in
MFT and other characteristics of the file are there.
If the original name more name first node resident of the index, then compare it
with the name of the second node, and if the original name below, the described
procedure is the second resident group names, etc. As a result, instead perebora
many names (in the worst case, all the name directory) is compared with a much
smaller number of nodes and names names in a group directory. If one group
becomes too large directory, it is also divided into groups, each of the last
names of the new team left in the original nerezidentnom attribute Index Root,
and all the other names on new teams moving into new non-resident type
attributes Index Root (in the figure did not show ). By starting nerezidentnomu
attribute Index Root added attribute of the index, indicating sections index of
new groups. If you now to find the desired name in the resident part of the
index level a comparison shows that the original name is less than one name kept
there, it is said that this attribute accurate comparison is to be named could
not be turning to the sub names next level tree.
File operations Two ways of file operations File System OS should allow users
operations set of files, designed in the form of system calls. This usually
consists of a set of system calls, as creat (a file), read (read from a file),
write (write to a file) and some others. Most of the same file the user is not
one operation, a sequence of operations. For example, when a text editor, file,
which contains a document, the user typically reads a few pages of text, editing
the data and write them on a matter and then reads a page from another file, etc.
After many operations reading and writing user completes work with the file and
moves to another. Whatever operations are not performed on the file, the OS must
perform certain universal for all operations of action: 1. On the symbolic name
of the file found its properties, which are stored in the file system on the
disk. 2. Copy features file into RAM, as the only way code can be used. 3. Based
on the characteristics of the file to verify the user to perform the requested
operation (read, write, delete, view file attributes). 4. Clear the area of
memory reserved under the temporary storage of files. In addition, each
transaction includes some unique for its action, such as reading a set of
clusters disk file management, etc. The operating system can perform the steps
of the file in two ways (Figure 7.26): For each operation performed as
universal or unique action. This is sometimes called the scheme without memory
of Operations (stateless). All of universal performed at the beginning and end
of the sequence of operations, and for each sub-operations are performed only
unique action. The vast majority of file systems supported a second method of
file operations as a more economical and faster. The first method has one
advantage-it is more resistant to breakdown in the system, because every
operation is self-sufficient and not dependent on the previous result. Therefore,
the first method is sometimes used in distributed network file system (for
example, Network File System, Sun NFS), which disrupted because of the loss of
packets or failure of a network nodes more likely than local access to the files.
|