Operating Systems

1
Operating Systems

• Files II - Implementation
• 11/20/00

2
File Implementation - Low Level

• Sequential vs. random access storage devices
• Secondary storage devices have
• volume directories
• external file descriptors (for each file)
• the files
• Block management
• assigning blocks to files
• keeping track of free blocks

3
Opening a file

• Locate file descriptor in file system
• Extract info about file from file descriptor and
  info about process from process descriptor
• Verify that process may access file
• Create an entry in the file status table
• Allocates I/O buffers (and other resources) to
  handle file usage

4
Closing a file

• Complete pending operations
• Release I/O buffers
• Release any locks the process holds on the file
• Updates the external file descriptor
• Deallocates the file status table entry

5
(No Transcript)
6
(No Transcript)
7
Contiguous Allocation

• Random device behaves like a sequential one
• Sequential blocks may be accessed quickly (little
  head movement on the device)
• May not have dynamic file sizes
• Tends to promote external disk fragmentation

8
(No Transcript)
9
Doubly linked lists - to improve performance
10
(No Transcript)
11
Unix
12
Unallocated blocks

• System uses a free list to monitor/keep track of
• Use a linked list or an indexed list
• Usually very large, so indexed list not good
• Using a ll, not easy to allocate disk blocks in
  the same area
• Sometimes use a block status map (disk bitmap) --
  the ith entry is set if the ith block is
  allocated
• This approach has the same problems bitmaps had
  when dealing with memory

13
Reading/Writing blocks

• 2 steps
• reading/writing to/from memory copy of block
• reading/writing memory lt---gtstorage devices
• Reading (unpacking) -- converting secondary
  storage blocks into a byte stream
• Writing (packing) -- moving bytes into copies of
  secondary storage, and writing them to secondary
  storage blocks when they become full

14
Tricky operations

• Insertion/Deletion
• Seeking

15
(No Transcript)
16
Block I/O

• For performance reasons, block buffering may be
  used
• Information is generally accessed sequentially

17
Other levels of file implementation

• Structured sequential files -- a full record is
  read at a time
• Indexed sequential files
• multiple records in 1 block
• each block gets its own record
• buffering doesnt help

18
Other levels (contd)

• DBMSs
• use a storage manager instead of a file manager
• specialized strategies for different DBs
• Multimedia
• documents generally implemented via abstract data
  types
• high bandwidth throughput
• multiple blocks requested at once

19
Memory mapped files

• File access is more complicated than regular
  memory acess
• Some OSs allow mapping of files into a running
  process AS
• MAP, UNMAP
• file becomes the backing for the file
• works best in a system that supports segmentation

20
(No Transcript)
21
Problems

• Hard to know exact lengths of output files
• Inconsistencies if file is opened for writing by
  1 process, but conventional read by another
• File may be larger than the segment