[libvirt] Re: [PATCH] also allow use of XZ for Qemu image compression

[Charles Duffy]

[Pardon the repost -- fixed the table formatting in this version]

Jim Meyering wrote:
> Good point about it being one of the fastest.
> I shouldn't have mentioned the subjective "popular".
> Usefulness trumps that.  I suppose Daniel, Cc'd, will decide.

Per off-list discussion with DV, I'm providing some numbers. Sort order
is space used on disk, lowest to highest, with a 251MB file created by
virDomainSave.

                 comptime[*]     decomptime[*]   space
xz[default]     3m45.890s       4.960s          42MB
xz[0]           0m17.450s       6.080s          54MB
bzip2           0m56.290s       11.080s         58MB
gzip            0m13.790s       2.260s          64MB
lzop            0m1.970s        0.800s          87MB

I believe this makes the distinct niche of each of the compressors
clear, with the exclusion of bzip2 (which is both slower and produces
larger output compared to xz -0).


[*] "time" is user-mode CPU time, as reported by the bash "time" builtin
on an Intel Xeon E7330 @ 2.40GHz.