yeah, realised that the colinux had the freenx
advantage in the look&feel dept. I agree its a mixed
bag. Took me a while to understand that there was
nothing to be done for networking under qemu, its just
that ping doesn't work. Although, I still don't know
how to ssh into the guest, because there is no way for
me to identify it. I like the colinux networking way
better and easier.
--- Doctor Bill <docbill@...> wrote:
> On 9/18/05, Sunil <funtoos@...> wrote:
> > Look and feel performance wise colinux is way
> > e.g. under qemu mouse jumps.
> Are you comparing apples to apples or apples to
> oranges. QEMU
> emulates a video card and output, coLinux doesn't.
> With coLinux you
> need to run an X server, or use VNC. Under those
> conditions you will
> never see the mouse jump. To be fair, connect to
> QEMU the same way
> you connect to coLinux, and minimize the emulated
> display. (I don't
> believe you can turn the emulated display off, but
> you can make sure
> it is not used durring your test.)
> > Colinux beats kqemu usability wise as well e.g. I
> > reusing the existing Gentoo install that I had on
> > another disk under colinux without much work,
> > qemu will need an image to be created.
> True, but QEMU can use the linux kernel you already
> have installed.
> So there are advantages both ways.
> > Networking was bigger hassle with qemu e.g. I
> > never make it talk to my gateway 192.168.0.1, it
> > always insisted on going thru its own aliases. I
> > haven't figured how do I get incoming connections.
> > Although, with colinux networking is not very
> > either.
> I found just the reverse. With QEMU, slirp in
> enabled be default, so
> you can connect to the internet from most Linux
> distributions with
> minimal effort. With coLinux, it takes a little
> effort to enable
> slirp. Personally, I prefere slirp because I
> normally don't want
> guest machines to have their own IP address on the
> subnet. But I
> suppose it depends on what you use the guest OS for.
> > I did some quick simple build tests.
> > mrxvt-0.4.1: colinux = 1m22secs, kqemu = 3m21secs
> > fontconfig-2.3.2: colinux 1m54secs, kqemu =
> Clock timings are unreliable for both coLinux and
> qemu. For an
> accurate timing comparison, pick tests you can time
> with a stop watch.
> Based on my experience, your results look
> reasonable. However, I
> would suggest trying tests that are not as IO
> intensive as builds.
> What image file format did you use?
> > That's more than twice as fast!
> Yes. That is the predicted performance difference.
> > The restful CPU usage was always around 10% for
> > while it was around 4% for colinux.
> Hmmm. My rest CPU usage for coLinux is usually 0%.
> I never tried
> measuring it with qemu because, I don't expect qemu
> to be resource
> friendly even with kqemu.
> > Its all colinux in all departments, all the way!!
> Definitely not all departments. For ease of setup
> of a new Linux
> distrubution, qemu is much better. For performance
> coLinux is better.
> For networking it is a tossup depending on your
> needs. For platform
> portability, qemu is better. For open source
> projects, coLinux is
> better since the kqemu accelerator is not open
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around