第6章インフラストラクチャーの抽象化層：オペレーティングシステム

By the end of this section, you will be able to:

• Explain features of various file systems

• Discuss file system structures and layers

In this module, we learn about files, file management, disk devices, file systems, file system interface, and distributed file systems.

A file is a collection of related information that is stored on a storage instrument such as a disk or secondary/ virtual storage. It is the smallest storage unit from the user’s perspective. The file name includes two parts: name and extension (e.g., filename.txt). Each extension is for a specific purpose such as .exe (in Windows OS) to run a program and .txt for text files.

A file system is responsible for defining file names, storing files on a storage device, and retrieving files from a storage device. When designing a file system for managing many files, some issues to consider are as follows: most files are small so per-file overheads must be low; most of the disk space is in large files; many of the I/O operations are for large files so performance must be good for large files; files may grow unpredictably over time; users want to use text names to refer to files.

Special disk structures called directories are used to map names to support hierarchical directory structures. A directory is a set of files that is managed by the OS, and it also contains all the required information about the files, such as attributes, location, and ownership. The UNIX/Linux approach is as follows: directories are stored on disk just like regular files except with extra information to indicate that it is a directory. Each directory contains <name, address> pairs. The file referred to by the address may be another directory; hence, we can have nested and hierarchical directory structures.

The problems facing modern file systems include disk management, naming, and protection. File systems are often trying to improve access to files by minimizing seeks, sharing space between users, and making efficient use of disk space. A system’s ability to reduce faults and ensure that the information in the system survives OS crashes and hardware failures is called its reliability. In addition to improving reliability, a file system should guarantee a high level of protection by maintaining isolation between users and controlling the sharing of resources.

While file systems are a layer of abstraction that provides structured storage and defines logical objects such as files and directories, disk devices are considered raw storage. Data that can be directly accessed by the CPU with minimum or no delay and does not survive a power failure is held in primary storage. Persistent memory that survives power failures most of the time, such as spinning disks, SSDs, and USB drives, is considered secondary storage. Routines that interact with disks are typically at a very low level in the OS and are used by many components such as file systems and virtual machines. These secondary storage devices may handle the scheduling of disk operations, error handling, and often the management of space on disks. A trend is for disks to do more of this themselves.

Operating systems use various methods to locate files by their names, and the methodology often depends on their underlying file system architecture. To illustrate these concepts, here are some examples from UNIX-like systems and Windows:

• UNIX/Linux (Inodes): In UNIX-like systems, the file system uses a structure called an inode to represent files and directories. An inode contains metadata about a file or directory but not its name. The name-to-inode mapping is stored in directories, which are special files that list names of files and their corresponding inodes. When searching for a file by name, the OS starts at the root directory and follows the path specified in the file name. Each part of the path is looked up in the current directory’s list of names and their associated inodes. The OS reads the directory file, finds the name, and retrieves the inode number, which then leads to the inode itself. The inode provides the location of the data blocks, allowing the OS to access the file’s data. This process may involve multiple steps if the file is in a nested directory structure.

• Windows (File Allocation Table and NTFS): In File Allocation Table (FAT) format, files are located using a table that maps file names to the clusters (blocks) on the disk where their data is stored. The FAT is essentially a list, with each entry containing the location of the next part of the file. This creates a chain that the OS follows to read the entire file. In New Technology File System (NTFS), files are located using a Master File Table (MFT), and each file and directory on an NTFS volume has an entry in the MFT containing data, including the file name, size, time stamps, permissions, and the locations of the file’s data on disk. When searching for a file, the OS consults the MFT to find the entry corresponding to the file name, which then provides the information necessary to access the file’s data.

As you may recall, files represent values stored on disk and directories represent file metadata. File systems define operations on objects such as create, read, and write, and they may also provide higher-level services such as accounting and quotas, incremental backup indexing or search, file versioning, and encryption. A quota is the amount of space to store files based on the available memory space. Quotas are used to protect the system from unnecessary load and help in organizing the data in the storage. An incremental backup is a backup image containing the pages that have been updated from the time of the previous backup. The method that converts the data into secret code that hides the data’s true meaning is called encryption. The system that allows a file to exist in several versions at the same time, which gives the user complete control over file creation as in the file versioning example, is called file versioning (Figure 6.33).

File systems are concerned with lower-level characteristics such as performance and failure resilience. Both performance and failure resilience may be strongly affected by hardware characteristics.