- This topic has 4 replies, 3 voices, and was last updated Dec 11-7:54 pm by BitJam.
-
Posts
-
Xaver
The default value for xz compression is “-6”. The strongest (and slowest) setting is -9. Is it possible to set xz compression in iso-snapshot to “-9”? I have tried to do that via environment variable in .profile and /etc/bash.bashrc.
XZ_OPT=-9 (also tried XZ_OPT="-9", XZ_OPT=-1, XZ_OPT="-1") export XZ_OPTBut the compression rate did not change, no matter which value I have chosen. Is there a different solution?
Anonymous
Doing so yields a grossly inefficient result (discovered when I tested a while back)
but if you care to find out firsthand, to see for yerself…edit the configuration file /etc/iso-snapshot.conf
and therein declare your desired options.
isosnapshot parses the mksq_opt line and, without validating or manipulating your declaration, passes the string to mksquashfs. No extra-quoting of values should be necessary; just use the same syntax as you would when directly calling mksquashfs.Before you dive in, I recommend that you first read
https://jonathancarter.org/2015/04/06/squashfs-performance-testing/In case you don’t recognize the name, Jonathan Carter was a 2019 Debian Project Leader candidate.
Articles throughout his blog at jonathancarter.org are quite informative, as well as his instructional videos
playlist (60+ videos): Debian Package of the DayXaver
@ skidoo – Thank you
Does a higher compression level simply mean to get a larger block size? ––>
-6 = block size 131072
-7 = block size 262144
-8 = block size 524288
-9 = block size 1048576Then the following setting would give me xz compression with level -9:
mksq_opt=-comp xz -b 1048576Is that true?
According to Jonathan Carter level 9 would be about 8% smaller than level 6 and would take about 28% longer to be compressed.
The larger block size would also slow down the live system.“Is that true?”
Yes it is, according to the mksquashfs manpage.
You might also wish to investigate the effect from tweaking the mksquashfs -Xdict-size option.
BitJam
There are two different block sizes. I believe the mksquashfs block size is the size of the different compressed chunks inside the file. If the entire file were compressed in one chunk (like in in a compressed tar file) then most files would take ages to read. Squashfs gets its speed by doing the compression in chunks. A larger blocksize gives you better compression but could slow down random access. As skidoo mentioned, the mksquashfs -Xdict-size option controls the xz dictionary size which seems to be equivalent to controlling the xz compression. But the xz man page warns:
“The selected compression settings determine the memory requirements of the decompressor, thus using a too high preset level might make it painful to decompress the file on an old system with little RAM. Specifically, it’s not a good idea to blindly use -9 for everything like it often is with gzip and bzip2.”
If you can find a way to get better compression without sacrificing decompression speed, that would be awesome!
Context is worth 80 IQ points -- Alan Kay
- You must be logged in to reply to this topic.
