Return to BSD News archive
#! rnews 3113 sserve.cc.adfa.oz.au
Path: sserve!newshost.anu.edu.au!harbinger.cc.monash.edu.au!msunews!uchinews!vixen.cso.uiuc.edu!howland.reston.ans.net!swrinde!gatech!news.mathworks.com!news.kei.com!nntp.et.byu.edu!news.byu.edu!hamblin.math.byu.edu!park.uvsc.edu!usenet
From: Terry Lambert <terry@cs.weber.edu>
Newsgroups: comp.unix.bsd.bsdi.misc
Subject: Re: [Q]information about Berke Us
Date: 24 May 1995 22:31:39 GMT
Organization: Utah Valley State College, Orem, Utah
Lines: 59
Message-ID: <3q0c4b$8mc@park.uvsc.edu>
References: <3pudsb$oai@jagalchi.cc.pusan.ac.kr>
NNTP-Posting-Host: hecate.artisoft.com
kimjh@hyowon.cc.pusan.ac.kr (Jonghyun Kim) wrote:
]
] I would like to know about Berkely fast file system(UNIX FFS).
]
] * difference between UNIX FFS and current UNIX BSD.
The current BSD's all implement the "UNIX" (Actually, the
*Berkeley*) Fast File System (FFS). It is the default file
system type, although others are available (like LFS, a Log
Structured File System, largely out of the Berkeley SPRITE
project).
The difference is the file system framework and the actual code
in the implementation. Other than the "clean bit" extension to
reduce startup time after a clean shutdown, the 64 bit support,
the use of negative block addresses for indirect blocks, and
the symboliclinks being pushed into the directories (all of
which are described in the release notes) there are no real
differences. Most of these could be described as implementation
details.
> * how much popular UNIX FFS ?
It's the default on most systems, except AIX (which owes some of
it's JFS design to it, and Veritas, the UnixWare default file
system (UFS [FFS] *is* optionally usable as a replacement). The
Veritas file system contains large chunks of UFS/FFS in it, with
major emphasis on directory structures, etc. (some of the code
is directly derived from USL's UFS/FFS implementation).
[ I'll assume you meant to type "FFS", not "FFT" in the following ]
> * costs of UNIX FFT
Generally "Free with the system". UFS and FFS are much the same
thing. The BSD4.4 framework distinguishes them for file type
purity.
> * performance of UNIX FFT compared to current UNIX BSD
UFS/FFS is equally performant. For files with more than 0 layers
of block indirection, the use of negative block addresses will save
one block fault dor 1st indirects. This is a minimal savings and
would probably be a cache hit as opposed to a read in typical
usage. The Heidemann Ficus/BSD4.4 stacking framework adds some
minimal overhead.
UFS on UnixWare with an equivalent (8k) block size is within 3%
of the Veritas numbers (beating them for some operations). It
*seems* that the UFS block size was intentionally crippled at a
4k default in UnixWare releases to make VXFS look much better than
it was, and a correction of an async write of metadata information
makes it nearly as reliable in the face of a power failure.
Terry Lambert
terry@cs.weber.edu
---
Any opinions in this posting are my own and not those of my present
or previous employers.