[oi-dev] seeking a modestly brave soul to run some basic benchmarks

[Bayard Bell]

After seeing OI/Illumos on VBox suck wind compiling, I've had a chat with some of their developers on #vbox-dev about the problems I'm seeing. They're receptive and simply want to see reproducible simple test cases. Their preferred case is the time to compile VBox OSE itself on a guest. I ran a simple benchmark: compare a guest with 2 vCPUs and 2GB to a guest with 4 vCPUs and 4 GB. The hosts are bloomington and indianapolis, respectively. I ran the compilations twice, with a clean target between the two, just to see if there was any improvement on a second run. Here's what I got from timex:

first run result on blooomington                                                                                                                                                                                                                                                                                               

real       27:26.03
user       31:39.80
sys        11:24.31

second run results on bloomington

real       37:08.96
user       36:17.43
sys        17:06.48

first run result on indianapolis

real       58:50.34
user     1:29:07.86
sys      1:21:58.30

second run result on indianapolis

real     1:03:52.51
user     1:41:24.83
sys      1:47:03.53

The degradation going from 2 to 4 vCPUs is consistent with what I've seen previously, and the VBox devs agree that this looks off. Their only concern before proceeding is that my host OS is OS X, which is the least performant and supported host OS for VBox. They'd simply like to see the results reproduced on a 64-bit Solaris, Linux or Windows host with a vaguely comparable hardware baseline (I've got 2x4-core 2.8GHz CPUs with 24GB memory: I'm sure you can do with as little as 8GB if the system isn't running anything else). My guest install is freshly updated pkg.oi.o/dev-il, upgraded from oi-dev-148-text-x86.iso. The build instructions have been updated in response to my queries about dependency information:

http://www.virtualbox.org/wiki/Solaris%20build%20instructions

The only correction I'd make is that the configure line for Qt4 should be (IIRC):

./configure -v -platform solaris-g++-64 -shared -stl -largefile -sm -qt-libjpeg -qt-libpng -qt-libmng -qt-zlib \
    -prefix /opt/vboxose-qt \
    -I/usr/include \
    -I/usr/X11/include \
    -I/usr/X11/share/include \
    -I/usr/sfw/include
And the configure line for VBox itself should be:

./configure --disable-hardening --with-qt-dir=/opt/vboxose-qt
Even if people have no interest in using VBox themselves, please consider that people will be trying out OI on VBox and that they are not likely to adopt if what seems like it should be a well-provisioned guest performs as though hog-tied.

Also, please feel free to alert me to anything I've missed up to now.

Cheers,
Bayard
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://openindiana.org/pipermail/oi-dev/attachments/20110526/d963df0c/attachment-0004.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 1515 bytes
Desc: not available
URL: <http://openindiana.org/pipermail/oi-dev/attachments/20110526/d963df0c/attachment-0008.bin>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: PGP.sig
Type: application/pgp-signature
Size: 841 bytes
Desc: This is a digitally signed message part
URL: <http://openindiana.org/pipermail/oi-dev/attachments/20110526/d963df0c/attachment-0009.bin>