KVM performance problems (potentially a newbie issue)

[db_404]

I'm _think_ getting some significant kvm performance issues, my 'usual' virtualization environment is Xen, but I was wanting to try out kvm for it's newer kernel support and also for greater flexibillity.

The problem is that a seemingly low cpu load (say 20%) on the kvm guest, results in much higer (150%+) cpu loads on the guest, and from a user perspective the guest just 'feels' much slower than the host even though they are both non-graphical environments (this is a small server). This seems wrong and has me wondering if I've configured something in a fundamentally incorrect way. I'm wondering if it's just falling back to quemu emulation for the CPU - but I'm not sure how to check for this

My guest os is 2.6.27-gentoo-r8, with KVM paravirt options, and using paravirt network and block drivers, the host os is 2.6.28-gentoo-r4. KVM version is kvm-84.

In terms of CPU, it's an AMD dual athlon 64, so it has svm support:

Code: Select all

$ cat /proc/cpuinfo 
processor	: 0
vendor_id	: AuthenticAMD
cpu family	: 15
model		: 107
model name	: AMD Athlon(tm) Dual Core Processor 5050e
stepping	: 2
cpu MHz		: 2600.000
cache size	: 512 KB
physical id	: 0
siblings	: 2
core id		: 0
cpu cores	: 2
apicid		: 0
initial apicid	: 0
fpu		: yes
fpu_exception	: yes
cpuid level	: 1
wp		: yes
flags		: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt rdtscp lm 3dnowext 3dnow rep_good pni cx16 lahf_lm cmp_legacy svm extapic cr8_legacy 3dnowprefetch
bogomips	: 5224.31
TLB size	: 1024 4K pages
clflush size	: 64
cache_alignment	: 64
address sizes	: 40 bits physical, 48 bits virtual
power management: ts fid vid ttp tm stc 100mhzsteps

processor	: 1
vendor_id	: AuthenticAMD
cpu family	: 15
model		: 107
model name	: AMD Athlon(tm) Dual Core Processor 5050e
stepping	: 2
cpu MHz		: 2600.000
cache size	: 512 KB
physical id	: 0
siblings	: 2
core id		: 1
cpu cores	: 2
apicid		: 1
initial apicid	: 1
fpu		: yes
fpu_exception	: yes
cpuid level	: 1
wp		: yes
flags		: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt rdtscp lm 3dnowext 3dnow rep_good pni cx16 lahf_lm cmp_legacy svm extapic cr8_legacy 3dnowprefetch
bogomips	: 5211.31
TLB size	: 1024 4K pages
clflush size	: 64
cache_alignment	: 64
address sizes	: 40 bits physical, 48 bits virtual
power management: ts fid vid ttp tm stc 100mhzsteps

The kvm_amd module is loaded:

Code: Select all

# lsmod
Module                  Size  Used by
bridge                 53856  0 
stp                     6724  1 bridge
llc                    10016  2 bridge,stp
tun                    15684  2 
kvm_amd                34124  2 
kvm                   144048  1 kvm_amd

I'm starting my VM with a command like (using lvm for block storage):

Code: Select all

kvm -smp 2 -m 2048 \
-net nic,model=virtio,macaddr=52:54:00:12:34:01 \
-net tap -drive index=0,file=/dev/storage1/host-boot,if=virtio,boot=on \
-drive index=1,file=/dev/storage1/host,if=virtio \
-drive index=2,file=/dev/storage1/host-swap,if=virtio \
-vnc :1

Any thoughts as to why I'm seeing issues would be greatly appreciated.

[redbottle]

I don't see anything wrong with your setup, but the kvm faq has an entry on how to check if you're falling back to qemu mode.

[bbgermany]

Try without smp and you will see an incredible performance jump. The smp implementation of kvm seems to be a bit buggy. Search the forums and you will find some posts about the same issue.

bb

[db_404]

Thanks for the tip on smp, I'll give that a try, shame though, I could use the extra CPU horsepower.

I checked the qemu fallback information in the KVM faq, and I'm not falling back to quemu, so it may well be a smp issue.