Advanced Operating Systems   COMP9242 2002/S2

UNSW

 Printer-Friendly Version

Administration               
- Notices
- Course Intro
- Consultations
# On-line Survey (closed)
- Survey Results
 
Work
- Lectures
- Milestone 0
- Project Admin
- Project Spec
- Project FAQ
- Exam
 
Documentation
- ASysT Lab
- L4 source browser
- Sulima ISA Simulator
- R4x00 ISA Summary 
- MIPS R4700 Reference - MIPS R4000 User Manual 
- Network Driver
- GT64111
 
Related Info
- Aurema OS Prize
- OS Hall of Fame
 
History
- 2000
- 1999
- 1998
 
Staff
- Gernot Heiser (LiC)

 

Microsoft Windows-NT File System (NTFS)

Designed as a replacement of DOS FAT and OS/2 HPFS[Cus94].

• Requirements:
  • large files and devices ( 2GB),
  • protection,
  • reliability and recoverability.
• Features:
  • transaction semantics on metadata updates,
  • support for mirroring and striping,
  • Unicode file names,
  • secondary indices (not implemented),
  • extensible attribute set, incl.
    • multiple data streams,
  • unified access to all attributes (incl. data),
  • POSIX features: case-sensitive names, creation time stamp, hard links.
  • Many others every non-PC user expects from a file system.

File system meta-metadata:

The master file table (MFT).

• ``Relational database'' of file metadata:
  • transaction semantics,
  • secondary indices for accessing files (only implemented on names),
  • indices are B trees,
  • extensible on a per-record basis.
• One entry per file.
• Everything on disk is a file, including all metadata.
• Fixed length record (1-4kB) of variable-length attributes, incl.:
  • all names of each file (hard links, DOS name),
  • ACL
  • data (for small files ore directories).
• Entries are mostly one record, but can span several.

• Each attribute is represented by header and value.
• Value can be:
  • resident, i.e., in-line, or
  • non-resident, i.e., in separate file.
• Any value can be resident.
• Any value that can grow can be non-resident, incl.:
  • data,
  • file index (for directory),
  • ACL,
  • attribute list.
• Non-resident values
  • represented as array of:   (logical block #, physical block #, # blocks).
  • Supports sparse files (if compression attribute is on).
  • Also supports compression of data in extents.

Indices

• Directories are indices of file names, containing:
  • file number (= MFT record number),
  • time stamps (duplicate of MFT value),
  • size (duplicate of MFT value).
• All file names are also contained in MFT.
• Indices are B trees, represented as:
  • root node of names (key values),
  • corresponding child node (extent) pointers,
  • bitmap indicating which blocks in child extents are in use.

NTFS metadata

• All metadata is in files, described by the first 16 records of the MFT:
  • MFT,
  • partial copy of MFT (for fault tolerance),
    • mirrors entries for metadata files,
    • is allocated in the middle of the disk;
  • log file,
  • root directory,
  • boot file (at known address, contains address of MFT),
  • free block bitmap,
  • bad block file,
  • volume file (volume label),
  • ...

NTFS Recoverability

• Uses write-ahead logging for metadata updates.
  
  • Log contains redo and undo info.
  • Log entries are idempotent.
  • Log records may contain data or operations.
• Logging operation:
  
  1. Logs transaction sequentially in cache.
  2. Flushes log file.
  3. Updates metadata in cache.
  4. Writes commit record to log file.
• Occasionally writes checkpoint records to log:
  
  • Location recorded in (shadow-paged) log header.
  • Indicates starting point for recovery.
• When log full:
  
  1. stalls metadata update,
  2. flushes buffers,
  3. restarts log.
     

Micro$peak translation table

Cluster	Block
Volume	Partition
Run	Extent
Virtual cluster number	Logical block number
Logical cluster number	Physical block number