View previous topic :: View next topic |
Author |
Message |
Need4Speed Guru
Joined: 06 Jun 2004 Posts: 497
|
Posted: Thu May 01, 2008 8:39 pm Post subject: Reiser4, gzip or lzo compression? |
|
|
After looking at the benchmarks of reiser4 with compression I can't resist upgrading to reiser4 anymore. I understand gzip provides better compression than lzo at the cost of cpu usage. What are your thoughts on gzip vs lzo on an average modern desktop? |
|
Back to top |
|
|
dusanc Apprentice
Joined: 19 Sep 2005 Posts: 248 Location: Serbia
|
Posted: Thu May 01, 2008 8:42 pm Post subject: |
|
|
I use LZO1 because Edward Shiskin recomended it to me cause it has slightly lower copmpression on some types of data, and much smaller CPU requirements. |
|
Back to top |
|
|
kernelOfTruth Watchman
Joined: 20 Dec 2005 Posts: 6111 Location: Vienna, Austria; Germany; hello world :)
|
|
Back to top |
|
|
Need4Speed Guru
Joined: 06 Jun 2004 Posts: 497
|
Posted: Thu May 01, 2008 8:56 pm Post subject: |
|
|
So does storing data like mp3's and movies work worse with compression or does it just not provide any benefits? |
|
Back to top |
|
|
kernelOfTruth Watchman
Joined: 20 Dec 2005 Posts: 6111 Location: Vienna, Austria; Germany; hello world :)
|
|
Back to top |
|
|
dusanc Apprentice
Joined: 19 Sep 2005 Posts: 248 Location: Serbia
|
Posted: Fri May 02, 2008 1:19 pm Post subject: |
|
|
Reiser4 evaluates on first use if data is compressable. If it isn't it doesn't try to compress it again.
It would be interesting if someone could post some numbers about "slower" cryptocompress r4 against plain one.
There shouldn't be any differences as speed bottleneck is later down the path AFAIK and cryptocompress should be same or better in all situations. |
|
Back to top |
|
|
kernelOfTruth Watchman
Joined: 20 Dec 2005 Posts: 6111 Location: Vienna, Austria; Germany; hello world :)
|
Posted: Fri May 02, 2008 1:22 pm Post subject: |
|
|
dusanc wrote: | Reiser4 evaluates on first use if data is compressable. If it isn't it doesn't try to compress it again.
It would be interesting if someone could post some numbers about "slower" cryptocompress r4 against plain one.
There shouldn't be any differences as speed bottleneck is later down the path AFAIK and cryptocompress should be same or better in all situations. |
it highly dependens how you set up the cryptcompress plugin during format
there is compressMode=conv, compressMode=latt, compressMode=force, compressMode=ultim
compressMode=force should give the worst performance (?)
if you set up Code: | mkfs.reiser4 -o create=ccreg40,compress=gzip1 | or Code: | mkfs.reiser4 -o create=ccreg40,compress=lzo1 | it shouldn't be much worse than default reiser4, perhaps even better
Quote: | compression_mode
Points to a compression mode plugin which defines some heuristic for data compression. Currently this is a set of handlers called by deflate manager in flush time. Properly assigned compression mode allows to reduce cpu and memory usage when handling incompressible data.
List of currently possible values (label/description):
*
"ultim", "Check ultimately". In this mode compression will be "turned off" forever as soon as deflate manager meet incompressible page cluster.
*
"latt", "Check on dynamic lattice". This is adaptive mode that manage special per-inode compression toggle. This toggle shows last compression experience: ->discard_hook() sets it to 0; ->accept_hook() sets it to 1. ->should_compress() first looks at this toggle. If it is 1, then ->should_compress() advises to compress current logical cluster. Otherwise ->should_compress() advises to compress it only in the case when its index belongs to the lattice of variable factor X (X = 2, 4, ..., 32) which get increased after each unsuccessful experience (successful one returns it back to minimal value).
*
"conv", "Convert to extent". This mode activates a special hook in reiser4 context of sys_write() (when writing from special offset in some write iteration) which tries to estimate whether a file is compressible by testing its first logical cluster (64K by default). Such evaluation is performed only once per file's life. If evaluation result is negative, then fragments will be converted to extents, and management will be passed to unix-file plugin. Back conversion does not take place. If evaluation result is positive, then file remains under cryptcompress plugin control, but compression mode plugin will be converted to "latt", i.e all future decisions about compressibility will be in the competence of the flush algorithm.
*
"force", "Compress everything". In this mode deflate manager is forced to compress every logical cluster (unless it is too small).
*
"none", "Do not compress". In this mode deflate manager does not invoke compression.
Default value is "conv". |
_________________ https://github.com/kernelOfTruth/ZFS-for-SystemRescueCD/tree/ZFS-for-SysRescCD-4.9.0
https://github.com/kernelOfTruth/pulseaudio-equalizer-ladspa
Hardcore Gentoo Linux user since 2004 |
|
Back to top |
|
|
dusanc Apprentice
Joined: 19 Sep 2005 Posts: 248 Location: Serbia
|
Posted: Fri May 02, 2008 1:27 pm Post subject: |
|
|
Ofcourse
I was talking about default behavior as people usually don't (and IMHO shouldn't )tinker too much with those options. |
|
Back to top |
|
|
MostAwesomeDude Guru
Joined: 12 Aug 2007 Posts: 373
|
Posted: Fri May 02, 2008 2:52 pm Post subject: |
|
|
I find that LZMA works wonders on /usr/portage . _________________ Don't believe the "n00b" under my name. |
|
Back to top |
|
|
Need4Speed Guru
Joined: 06 Jun 2004 Posts: 497
|
Posted: Fri May 02, 2008 3:19 pm Post subject: |
|
|
MostAwesomeDude wrote: | I find that LZMA works wonders on /usr/portage . |
Does standard reiser4 support LZMA? What's the format command?
I know gzip worked really well too. It decreased my ~600mb portage sparse file to ~90mb! |
|
Back to top |
|
|
energyman76b Advocate
Joined: 26 Mar 2003 Posts: 2048 Location: Germany
|
Posted: Fri May 02, 2008 7:01 pm Post subject: |
|
|
I tried both and the differences on a several gb /var (which includes portage) was not that big. So I'd say go with lzo _________________ Study finds stunning lack of racial, gender, and economic diversity among middle-class white males
I identify as a dirty penismensch. |
|
Back to top |
|
|
MostAwesomeDude Guru
Joined: 12 Aug 2007 Posts: 373
|
Posted: Fri May 02, 2008 9:09 pm Post subject: |
|
|
Need4Speed wrote: | MostAwesomeDude wrote: | I find that LZMA works wonders on /usr/portage . |
Does standard reiser4 support LZMA? What's the format command?
I know gzip worked really well too. It decreased my ~600mb portage sparse file to ~90mb! |
Nope, but squashfs does, and reduced my portage to ~35MB. _________________ Don't believe the "n00b" under my name. |
|
Back to top |
|
|
|