File Systems and I/O - Module 17

Goals of File System Design

Efficiency, Usability, Reliability

History of File Systems

Basic FS, Performance ( efficiency and locality ), optimize writes, scale and distribution

File

a named collection of related information recorded on secondary storage

data

what a user or application puts in the file

metadata

file attributes added and managed by the OS

directory

a list that provides names for files

hard link

a mapping form each name to a specific underlying file or directory

soft link

a mapping from a file name to another file name

file system metadata examples

file system type, number of blocks in the file system, file system permissions

where is FS metadata stored?

superblock

what two data structures can you use to track free space for data and metadata?

free list, bitmap

file layout goals

fase access rate, efficient access to large files, allow files to grow past their initial size, allow random and sequential access with reasonable efficiency

contiguous allocation

allocation method where the file system allocates a contiguous chunk of free blocks when it creates a file, using a free list, start location, and size

sectors

smallest unit the drive can read or write, usually 512 bytes long

blocks

the smallest unit which the software accesses the disk at, which is an integer number of contiguous sectors

I/o device hardware consists of :

bus, device port, controller, device

linked allocation

allocation method where file is stored as a linked list of blocks, not neccesarily contiguous in memory but connected using pointers. needs the file header and a pointer to each involved block

FAT (File Allocation Table)

a simple table that stores linked list information

FAT Disadvantages

poor random access, limited access control, no support for hard links, volume and file size are limited, no support for transactional updates

direct allocation

allocation method where files are stored in individual blocks, with the location information being stored entirely in the file header

Indexed allocation

allocation method where files are stored in individual blocks, and location information fo rthem is stored in index blocks

why would we want to add index blocks to direct allocation

simpler to grow files

multilevel indexed files

allocation method where files are stored as a fixed, asymmetric tree, with the data blocks as its leaves

Fast File System

A UNIX file system designed to improve performance over older systems by using larger block sizes, cylinder groups for locality, block fragmentation for efficient small file storage, and optimized metadata allocation to reduce disk seeks.

directories are also called…

folders!

directory

a collection of mappings from file name to file number

where can directories be modified

kernel mode only

make-it-work strategy

a directory strategy where one name space is given to the entire disk; if one user uses a name, no one else can

user-based strategy

a directory strategy where each user has a single separate directory; once a user uses a name, that user cannot reuse the name

computer memory hierarchy, fastest to slowest

processor registers, processor cache, random access memory, flash / USB memory, hard drives, tape backup

seek time

time to position head over track/cylinder

maximum

time to go from innermost to outermost track

rotation time

time fo the sector to rotate underneath the head

transfer tim

time to move the bytes from disk to memory

polling

a communication method where the OS keeps checking until the status of the I/O device is idle, then carries out operations

in polling, what happens if the device is busy when the OS has a request

the OS moves to another task, but it keeps the operation on its todo list and tries back

interrupts

a communication method where the device can interrupt the CPU when it completes an I/O operation; then, the OS can determine which device causes the interrupt, if the last command was an input operation, and then start the next operation for that device

DMA controller ( direct memory access )

a communication method where the CPU is interrupted only when the entire transfer of data is complete, instead of when each word is ready

NTFS

A file system used by Windows that supports large files, journaling for reliability, access control via permissions, file compression, encryption, and metadata-rich structures for better performance and security.