Search the FAQ Archives

3 - A - B - C - D - E - F - G - H - I - J - K - L - M
N - O - P - Q - R - S - T - U - V - W - X - Y - Z
faqs.org - Internet FAQ Archives

Comp.os.research: Frequently answered questions [1/3: l/m 13 Aug 1996]
Section - [3.1] Extent-based versus log-structured file systems

( Part1 - Part2 - Part3 - Single Page )
[ Usenet FAQs | Web FAQs | Documents | RFC Index | Counties ]


Top Document: Comp.os.research: Frequently answered questions [1/3: l/m 13 Aug 1996]
Previous Document: [3] File systems
Next Document: [4] Mobile and disconnected computing
See reader questions & answers on this topic! - Help others by sharing your knowledge
From: File systems

[92-11-18-10-57.53] [92-11-22-10-16.26] A general definition for a
log-structured storage system might be the following: logging is an
append-only storage semantics.  The unit of logging is the record.
Write accesses append records to the end of the log.  A log record may
become obsolete.  Useless records are compacted out of the log when
possible.  Other write accesses are forbidden.

An extent-based file system is another candicate for better filesystem
performance.  The approach used under QNX, for example, is to have
files exist as an array of extents on disk, where each is extent is as
many contiguous blocks as could be allocated at that location.  By
using a bitmap for space allocation, files can also grow `in-place',
if adjacent free space on disk exists.  This approach allows the unit
of disk space allocation to remain 512 bytes, which is also the
smallest unit of physical I/O.  The potential performance bottleneck
of this approach does not happen because the filesystem code passes
I/O requests to the disk drivers in units of extents, not 512 byte
blocks.  The filesystem also heuristically modifies the size of the
pre-read requests based on the historical access pattern to the
file.  This approach provides the performance advantages of larger
physical disk block sizes, without the wasted disk space that results
from unused portions of large blocks, or the complexity of trying to
allocate space from partially unused blocks.


User Contributions:

Comment about this article, ask questions, or add new information about this topic:

CAPTCHA




Top Document: Comp.os.research: Frequently answered questions [1/3: l/m 13 Aug 1996]
Previous Document: [3] File systems
Next Document: [4] Mobile and disconnected computing

Part1 - Part2 - Part3 - Single Page

[ Usenet FAQs | Web FAQs | Documents | RFC Index ]

Send corrections/additions to the FAQ Maintainer:
os-faq@cse.ucsc.edu





Last Update March 27 2014 @ 02:11 PM