Diagnosis and Performance Monitoring - Deploying and Managing a Cloud Infrastructure - page 128

attributes, or simply “data about data.” It contains various kinds of information that may
be unique to the file:
■ File size
■ Creation date
■ Owner or creator
■ Location on the storage medium
■ File type or filename extension
■ Dimensions for image files
■ Compression algorithm
■ Security and permissions
■ Number of lines for documents and text files
■ Location of image or video shot with a GPS-enabled device
■ Encoding scheme of image, audio, or video file
Those are just some of the more common metadata fields you might find are viewable
by the user. The file system will require more specific information about the file that it will
not show to the user. File metadata is a type of file itself, and the file system keeps metadata
with the file it points to or in separate directories or even mediums. The latter is more prac-
tical, especially for checking the integrity of files.
Metadata gives raw structure to storage capacity and consists of descriptions and pointers
that link multiple sectors in a disk into coherent files and then identifies those files. Metadata
is required for persistent storage, and to qualify for that use, the file system must be able to
maintain its metadata integrity despite unpredictable system crashes such as ones caused
by power interruptions and operating system failures. Such crashes often result in total
information loss from volatile memory (RAM). The information that is kept in nonvolatile
storage such as disks must always remain consistent enough to be usable in the deterministic
reconstruction of a coherent state of the file system; there must be no dangling pointers or
uninitialized spaces, no multiple pointers causing ambiguous ownership of resources, and no
missing pointers to live resources. This requires the proper sequencing of updates in small
on-disk metadata objects.
However, metadata updates, such as during file creation and block allocation, have
always been the source of performance, security, integrity, and availability problems in
most if not all file systems. As an example of the complexity and performance cost associ-
ated with metadata, look no further than your traditional old-fashioned library with thou-
sands of topics cataloged by the library's holdings database. The database at a large library
is a small library in its own right. It is tedious to create, update, and manage the library
database. Although searching takes just a few keystrokes, the time spent for search may
depend on the size of the database and quality of the hardware. Once it's found, the librar-
ian has information about the topic such as author, publisher, publication date, and even
edition. The database here is the metadata of the topic. Managing and updating the data-
base can be cumbersome and time consuming, just like maintaining file system metadata.