Slow Performance OSX Host, Ubuntu 9.10 Guest
Posted: 9. Dec 2009, 06:31
Hello,
I'm comparing performance on VirtualBox and VMWare Fusion on OSX 10.5.7. Since I'm a developer, I'm mostly interested in compile speeds, so that's what I've done my benchmarking with. I use Fox Toolkit a lot, so I make my benchmarks using that. Running "time make" on Fox gives me the following numbers:
VirtualBox 3.0.12:
real 17m4.547s
user 7m51.053s
sys 10m20.815s
VMWare Fusion 3.0.0:
real 9m32.342s
user 3m27.509s
sys 5m48.070s
In playing around with this, I noticed that disk I/O seemed to be the major bottleneck on the VirtualBox side. To Test, I did just a straight dd from /dev/zero to disk as someone on another forum suggested. My results are as follows:
time dd if=/dev/zero of=file bs=1M count=1024
VirtualBox 3.0.12:
real 0m26.4243s
user 0m0.016s
sys 0m3.568s
VMWare Fusion 3.0.0:
real 0m3.678s
user 0m0.016s
sys 0m2.784s
I tried using a fixed-size disk file (in VirtualBox) thinking that may help, but the increase in performance was negligible.
I really like VirtualBox, and Sun Microsystems for that matter for all they do for the community. I'd really like to use this product, but unless I can figure out why this is so slow, I think I'll have to use VMWare. I do feel like I must be doing something wrong, because there's not really a lot of talk on the internet about this. In fact, most postings have VirtualBox's performance on par or better than other VM systems.
For the record, here's the rest of the data that I can think may be relevant:
Host: Mac OS X 10.5.7. MacBook Pro 15", 2.54G Core2Duo, 4G Ram
VirtualBox Setup: 1024MB RAM, ACPI and IO APIC enabled. 2 CPU, PAE/NX enabled. Guest Additions Installed.
VMWare Setup: 1024MB RAM, 2 CPU
Note that switching between 1 and 2 CPUs and switching the other parameters showed negligible effect. I also tried a PAE kernel in Ubuntu with little effect.
If there's something I can try, please let me know.
Thanks,
Alan.
I'm comparing performance on VirtualBox and VMWare Fusion on OSX 10.5.7. Since I'm a developer, I'm mostly interested in compile speeds, so that's what I've done my benchmarking with. I use Fox Toolkit a lot, so I make my benchmarks using that. Running "time make" on Fox gives me the following numbers:
VirtualBox 3.0.12:
real 17m4.547s
user 7m51.053s
sys 10m20.815s
VMWare Fusion 3.0.0:
real 9m32.342s
user 3m27.509s
sys 5m48.070s
In playing around with this, I noticed that disk I/O seemed to be the major bottleneck on the VirtualBox side. To Test, I did just a straight dd from /dev/zero to disk as someone on another forum suggested. My results are as follows:
time dd if=/dev/zero of=file bs=1M count=1024
VirtualBox 3.0.12:
real 0m26.4243s
user 0m0.016s
sys 0m3.568s
VMWare Fusion 3.0.0:
real 0m3.678s
user 0m0.016s
sys 0m2.784s
I tried using a fixed-size disk file (in VirtualBox) thinking that may help, but the increase in performance was negligible.
I really like VirtualBox, and Sun Microsystems for that matter for all they do for the community. I'd really like to use this product, but unless I can figure out why this is so slow, I think I'll have to use VMWare. I do feel like I must be doing something wrong, because there's not really a lot of talk on the internet about this. In fact, most postings have VirtualBox's performance on par or better than other VM systems.
For the record, here's the rest of the data that I can think may be relevant:
Host: Mac OS X 10.5.7. MacBook Pro 15", 2.54G Core2Duo, 4G Ram
VirtualBox Setup: 1024MB RAM, ACPI and IO APIC enabled. 2 CPU, PAE/NX enabled. Guest Additions Installed.
VMWare Setup: 1024MB RAM, 2 CPU
Note that switching between 1 and 2 CPUs and switching the other parameters showed negligible effect. I also tried a PAE kernel in Ubuntu with little effect.
If there's something I can try, please let me know.
Thanks,
Alan.