31436 Systems Software

1
31436 Systems Software NetworksAUTUMN 2002

• Tutorial 8
• File Management

2
QUESTION 1

• What is the difference between a file and a
  directory on UNIX?
• Would you expect the file access functions to
  work on directories?

3
UNIX Files and Directories

• Both types are essentially streams of bytes.
• UNIX administers files and directories alike by
  their iNodes.
• But where a file can contain any type of data,
  directories only contain ltfilename inodegt entries
  a list of files.
• No direct read/write operations can be performed
  on directories use mkdir and rmdir instead.

4
QUESTION 2

• Some systems provide file sharing by allowing
  several users to access a single copy of the file
  simultaneously. Other systems provide multiple
  copies of the files. Discuss the advantages and
  disadvantages of each approach.

5
Single Copy of a File

• Updates to the file can be viewed instantly by
  everybody.
• Relatively simple to queue up all requests to
  write to the file.
• But increases demand for network bandwidth.
• Must be careful with file access permissions
• No redundancy if file gets corrupted.
• Saves disk space.

6
Multiple Copies of a File

• Editing and updating the file is quicker, as each
  user has their own local copy.
• Eases network load
• Provides some redundancy in case one copy of the
  file becomes corrupted.
• Becomes very difficult to maintain file
  consistency several versions of the same file
  are on separate workstations.
• Waste of disk space.

7
QUESTION 3

• What are the relative merits and drawbacks of
  each of these disk block allocation methods?
• Contiguous
• Chained
• Indexed

8
Contiguous Allocation

• Good performance from using sequential disk
  blocks.
• Easy to randomly retrieve a single block.
• Can be difficult sometimes to allocate contiguous
  blocks can result in external fragmentation.
• Pre-allocation demands file size declaration at
  the time of creation.
• Disk blocks can vary in size.
• One entry required in file allocation table.

9
Chained Allocation

• Poor performance is disk blocks arent close.
• No adherence to principle of locality.
• Difficult to randomly retrieve a single block.
• Easy to allocate disk blocks one at a time. No
  external fragmentation possible now.
• Suited to sequential files processed
  sequentially.
• No pre-allocation no worries!
• One entry required in file allocation table.
• Fixed-size disk blocks.
• Very difficult to reconstruct a file if one or
  more of its pointers has become corrupted.

10
Indexed Allocation

• Easy to allocate disk blocks one at a time. No
  external fragmentation possible now.
• File index is on leading block for that file.
• File index block is very vulnerable!
• Easy to randomly retrieve a single block.

11
QUESTION 4

• Consider a UNIX file system. Assume that there
  are 12 direct block pointers, and a singly, a
  doubly, and a triply indirect pointer in iNode.
  Further, assume that the system block size is
  8KB. If the disk block pointer is 32 bits, with 8
  bits to identify the physical disk and 24 bits to
  identify the physical block, then
• What is the maximum file size supported by this
  system?
• What is the maximum file system partition
  supported by this system?
• Assuming no information other that the file iNode
  is already in main memory, how many disk accesses
  are required to access the byte in position
  13,423,956?

12
iNodes

• The iNode in this example includes 60 bytes of
  address information that is organized as 15
  32-bit (4-byte) addresses, or pointers.
• The first 12 addresses are direct pointers.
• The 13th address is a single indirect pointer.
• The 14th address is a double indirect pointer.
• The 15th address is a triple indirect pointer.

13
Direct Pointers

• Direct pointers point straight to the first n
  blocks of the file.
• 12 direct pointers point to the first 12 blocks.
• Direct Pointer 0 Block 0 8KB
• Direct Pointer 1 Block 1 8KB
• Direct Pointer 11 Block 11 8KB
• TOTAL 96KB

14
Single Indirect Pointers

• System block size 8KB 8192 Bytes
• Block Pointer 32 bits (32/8) Bytes 4 Bytes
• Block OF Pointers 8192/4 2048 pointers per
  block
• SID Pointer SID Block Block 0 8KB
• Block 1 8KB
• Block 2047 8KB
• TOTAL 16,384KB
• 16MB

15
Double Indirect Pointers

• 2048 pointers
• DID Pointer DID Block SID Block 0 Block 0 8KB
• Block 2047 8KB
• SID Block 2047 Block 4192255 8KB
• Block 4194303 8KB
• TOTAL 33,554,432KB
• 32,768MB
• 32GB

16
Triple Indirect Pointers

• 2048 pointers 2048 pointers
• TID Pointer TID Block DID Block 0 SID
  Block 0 Block 0 8KB
• Block 2047 8KB
• SID Block 2047 Block 4192255 8KB
• Block 4194303 8KB
• DID Block 2047 SID Block 4192255 Block
  8585738239 8KB
• Block 8585740287 8KB
• SID Block 4194303 Block 8589932543 8KB
• Block 8589934591 8KB
• TOTAL 68,719,476,736KB
• 64TB

17
Maximum File Size

• Add up maximum supported size of all components
  of the iNode.
• 12 direct pointers 96 kilobytes
• Single indirect pointer 16 megabytes
• Double indirect pointer 32 gigabytes
• Triple indirect pointer 64 terabytes
• 96KB 16MB 32GB 64TB

18
Maximum Partition Size

• oi

19
Disk Accesses

• We want byte no. 13,423,956 ? 13,109KB.
• The iNode is already in memory.
• Thus the 15 pointers contained in this iNode are
  already in memory and doesnt require a disk
  access.
• First 12 direct pointers cover first 96KB.
• The first single indirect pointer then covers the
  next 16,384KB which covers KB no. 13,109.
• So two access are required first access is the
  first SID, the second access is the block itself.

20
Need help? Ask me! )

• Daniel COMARMOND
• E-Mail dacomarm_at_it.uts.edu.au
• ICQ 6606132
• Phone [phone]
• Website http//charlie.it.uts.edu.au/dacomarm
• THE SECRET WORD
• Username www
• Password rosebud