3 . 5
W I N D O W S N T ' S N T F S
41
attributes may have only their header stored in the i-node, and they only
contain pointers to their associated data.
If a file has too many attributes to fit in a single i-node, another attribute
is added, an attribute list attribute. The attribute list attribute contains the
i-node number of another slot in the MFT where the additional attributes can
be found. This allows files to have a potentially unbounded list of attributes.
NTFS stores file and attribute data in what it refers to as "attribute
streams." NTFS uses extents to record the blocks allocated to a file. Ex-
tents compactly refer to large amounts of disk space, although they do suffer
the disadvantage that finding a specific position in a file requires searching
through the entire list of extents to locate the one that covers the desired
position.
Because there is little information available about the details of NTFS, it
is not clear whether NTFS uses indirect blocks to access large amounts of file
data.
File System Metadata
NTFS takes an elegant approach toward storing and organizing its metadata
structures. All file system data structures in NTFS, including the MFT itself,
are stored as files, and all have entries in the MFT. The following nine items
are always the first nine entries in the MFT:
MFT
Partial MFT copy
Log file
Volume file
Attribute definition file
Root directory
Bitmap file
Boot file
Bad cluster file
NTFS also reserves eight more entries in the MFT for any additional sys-
tem files that might be needed in the future. Each of these entries is a regular
file with all the properties associated with a file.
By storing all file system metadata as a file, NTFS allows file system struc-
tures to grow dynamically. This is very powerful because it enables growing
items such as the volume bitmap, which implies that a volume could grow
simply by adding more storage and increasing the size of the volume bitmap
file. Another system capable of this is IBM's JFS.
NTFS stores the name of a volume and sundry other information global to
the volume in the volume file. The log is also stored in a file, which again en-
ables the log to increase in size if desired, potentially increasing the through-
put of the file system (at the cost of more lost data if there is a crash). The
Practical File System Design:The Be File System
, Dominic Giampaolo
page 41