home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.os.linux
- Path: sparky!uunet!munnari.oz.au!news.hawaii.edu!spectra!oles
- From: oles@spectra (Shawn Oles)
- Subject: Re: Staker like file system
- Message-ID: <1992Aug27.213458.1983@news.Hawaii.Edu>
- Sender: root@news.Hawaii.Edu (News Service)
- Nntp-Posting-Host: spectra.eng.hawaii.edu
- Organization: University of Hawaii, Dept. of Electrical Engineering
- References: <1992Aug22.223110.21426@news.Hawaii.Edu> <2735@nlsun1.oracle.nl>
- Date: Thu, 27 Aug 1992 21:34:58 GMT
- Lines: 50
-
- In article <2735@nlsun1.oracle.nl> jhelberg@nl.oracle.com (Joost Helberg) writes:
- >Compressed file-system: don't you have big problems with page-alignment,
- >a page on disk cannot possibly resemble a page in core! Alignment of all but
- >the first page read-in is lost too!
- >
- >How do you page?
- >
- >How can exec see how big the executable is, when it is not uncompressed yet?
- >(exec needs to know, it must know the number of page-table levels to
- >allocate, it must know whether the executable fits into core+swap).
- >
- >Do you swap compressed to /dev/swap, or not, then it is better to swap
- >text out too, otherwise swapping will slow down the system awfully!
- >
- the compressed file system will not allow swap files, but PAGING of executables
- and complete random access is implemented by compressing files at the block level.
- Of course, doing this limits the possible compression, but this does not
- bother me too much.
-
- Swaping could be implemented, but would not be safe, first because you would
- never be able to tell how much memory is available on the swap device because
- it will be changing as a factor of achieved compression, and secondly the
- performance of Paging memory to disk would really suck unless the paged memory
- contained highly compressed data, which would limit the disk accesses/page.
-
-
- >Good luck implementing this, but the performance can't be comfortable I guess.
-
- I realize that there is a lot of doubt as to whether or not a compressed file
- system will work out in UNIX. In fact, the main reason that I am working
- on this project is to verify this with an implementation. If ineeded it
- can be made to get "comfortable" performance though, many people would be
- interested.
-
- Right now I use a compression algorithm based on work done by
- Ziv-Lempel 's 1977 paper and fined tuned by Ross Williams. The compression
- is very fast, but limits attainable compression. After I finally release
- ACF, anyone will be able to add in a different compression algorithm if they
- want. Also, it might be possible to do the compression using the Stacker
- card, which would allow other programs to do things while compression is
- occuring.
-
- Right now ACF (Automatically Compressed Format) works as a simultation,
- but blows up when I mount it. I am having some trouble tracking down
- the problems, but hopefully should get it working soon.
-
- -shawn
- --
-
- -shawn
-