Note for vSRX-NG (18.4 tested)
Posted: Tue Jan 14, 2020 8:55 am
Hi
When creating nodes with Juniper vSRX (I am using 18.4R3.3) - I found the 'default' QEMU string is insufficient.
It's missing "+vmx". This seems to be the QEMU flag that tells the guest VM to utilise hardware acceleration. Without it, the Junos Windriver Linux boot message says something like "kvm: no hardware acceleration available" in the telnet terminal on boot. The performance by default is also terrible.
Here is the string I am using at the moment, all other parameters (QEMU version etc) are default, although I suspect we can improve things more with tweaking.
"-machine type=pc-1.0,accel=kvm -cpu qemu64,+vmx,+ssse3,+sse4.1,+sse4.2,+x2apic,+aes,pclmulqdq -serial mon:stdio -nographic -no-user-config -nodefaults -rtc base=utc"
This makes them perform almost as well as a real one
EVE-NG version: 2.0.3-102
When creating nodes with Juniper vSRX (I am using 18.4R3.3) - I found the 'default' QEMU string is insufficient.
It's missing "+vmx". This seems to be the QEMU flag that tells the guest VM to utilise hardware acceleration. Without it, the Junos Windriver Linux boot message says something like "kvm: no hardware acceleration available" in the telnet terminal on boot. The performance by default is also terrible.
Here is the string I am using at the moment, all other parameters (QEMU version etc) are default, although I suspect we can improve things more with tweaking.
"-machine type=pc-1.0,accel=kvm -cpu qemu64,+vmx,+ssse3,+sse4.1,+sse4.2,+x2apic,+aes,pclmulqdq -serial mon:stdio -nographic -no-user-config -nodefaults -rtc base=utc"
This makes them perform almost as well as a real one
EVE-NG version: 2.0.3-102