Skip to Content.
Sympa Menu

sm-discuss - Re: [SM-Discuss] SquashFS resource usage

sm-discuss AT lists.ibiblio.org

Subject: Public SourceMage Discussion List

List archive

Chronological Thread  
  • From: Ismael Luceno <ismael.luceno AT gmail.com>
  • To: David Kowis <dkowis AT shlrm.org>
  • Cc: sm-discuss AT lists.ibiblio.org
  • Subject: Re: [SM-Discuss] SquashFS resource usage
  • Date: Mon, 16 Jan 2012 03:49:38 -0200

On Sun, 15 Jan 2012 14:17:57 -0600
David Kowis <dkowis AT shlrm.org> wrote:
> What about performance hits regarding searching for spells, indexing,
> various gaze functions and whatnot.
>
> Surely there's a bit of a hit, but I'm not sure how significant it'd
> be.

The CPU and memory access overhead is, IMHO, mostly unnoticeable.

I think only the fastest SSD disks (+300MB/s), used in conjunction with
very low end CPUs would actually find decreased performance. Almost
everything else should favor compressed filesystems.

I've made a simple test. The file:
-rw-r--r-- 1 ismael users 127M Jan 16 01:49 testfile.tar.gz

So i tried decompression in RAM (tmpfs):
# time gzip -dc testfile.tar.gz |wc -c
702668800

real 0m6.691s
user 0m6.116s
sys 0m1.077s

Even with all the overhead of doing it in userspace, it reached
about 109MB/s for a file compressed with gzip -9. This machine has a
2.0GHz Athlon64 CPU and 1GB DDR400 memory, so it's quite slow for
nowadays standards.

Even on this system, it would require just a second to unpack the whole
grimoire.

This machine has a IDE disk that can put about 60MB/s at best, so it
would benefit a lot. But even with a modern SATA2 disk, it would give
between 70 and 100MB/s on an ideal scenario, and given it's easy to
get it all fragmented and that the latencies are huge, it would benefit
anyway.

It would be interesting to see more tests ;).

Attachment: signature.asc
Description: PGP signature




Archive powered by MHonArc 2.6.24.

Top of Page