02-05-2021, 03:11 PM
In this post we'll talk a bit about QEMU/KVM guest CPUs used in VirMach VPS-9(s) before ending with the right virt-install command that will approximate the CPU model used in Phoenix VPS-9.
To make the case, i'll use the already published data of 3 VPS-9 (Atlanta, L.A., Seattle.) I'll also use my own -still unpublished- data on VPS-9 @Phoenix and @Buffalo.
1-VPS-9 Used Guest CPUs:
From the available data, we have:
1.1- VPS 9 (Atlanta)
https://post4vps.com/Thread-Virmach-VPS-...-Dream-VPS
1.2- VPS 9 (L.A.)
https://post4vps.com/Thread-Virmach-VPS-9-Review
1.3- VPS 9 (Seattle)
https://post4vps.com/Thread-Virmach-VPS-...ew-Seattle
1.4- VPS 9 (Phoenix)
Unpublished__data
1.5- VPS 9 (Buffalo)
Unpublished__data
2- Libvirt CPU Models :
Libvirt supports three ways to configure guests CPU models:
Armed with the above information in mind, we're now in a position to fine-tune our virt-install command even further, like so:
Again, why is it important to add the cpu flag in the command?.. Answer, to avoid ending up with a host-model CPU (on my system, it's Model: IvyBridge-IBRS that resolve to : Intel Xeon E3-12xx v2 inside the guest) instead of the more powerful host-passthrough model that will resolve to your own host CPU (ie Intel® Core i5-3470 CPU @ 3.20GHz on my KVM system.)
In the next post we'll talk a bit about disk I/O fine-tuning.
To make the case, i'll use the already published data of 3 VPS-9 (Atlanta, L.A., Seattle.) I'll also use my own -still unpublished- data on VPS-9 @Phoenix and @Buffalo.
1-VPS-9 Used Guest CPUs:
From the available data, we have:
1.1- VPS 9 (Atlanta)
https://post4vps.com/Thread-Virmach-VPS-...-Dream-VPS
Code: (Select All)
Processor : QEMU Virtual CPU version (cpu64-rhel6)
CPU Cores : 2 @ 2499.998 MHz
Kernel : 3.10.0-1062.18.1.el7.x86_64
Geekbench 5.0.1 ( https://browser.geekbench.com/v5/cpu/2472702 )
-Single-Core Score : 316
-Multi-Core score : 414
System Benchmarks Index Score x1 250.8
System Benchmarks Index Score x2 348.4
1.2- VPS 9 (L.A.)
https://post4vps.com/Thread-Virmach-VPS-9-Review
Code: (Select All)
Processor : Intel Xeon E312xx (Sandy Bridge, IBRS update)
CPU cores : 2 @ 2499.998 MHz
Kernel : ?
Geekbench 5.3.1 ( https://browser.geekbench.com/v5/cpu/5391725 )
-Single-Core Score : 376
-Multi-Core score : 646
1.3- VPS 9 (Seattle)
https://post4vps.com/Thread-Virmach-VPS-...ew-Seattle
Code: (Select All)
Processor : Intel® Xeon® CPU E5-2670 v2 @ 2.50GHz
CPU Cores : 2@ 2499.998 MHz
Kernel : 3.10.0-1062.4.1.el7.x86_64
Geekbench 5.1.0 ( https://browser.geekbench.com/v5/cpu/2077291 )
-Single-Core Score : 387
-Multi-Core score : 702
System Benchmarks Index Score x2 664.6
1.4- VPS 9 (Phoenix)
Unpublished__data
Code: (Select All)
Processor : Intel(R) Xeon(R) CPU E5-2620 v3 @ 2.40GHz
CPU cores : 2 @ 2399.996 MHz
Kernel : 4.18.0-147.8.1.el8_1.x86_64
Scores for GeekBench 5.1.1, 5.1.0 and 4.3.1, in single and multi-core tests:
>> GeekBench 5.1.1 : 368 ; 572
>> GeekBench 5.1.0 : 392 ; 651
>> GeekBench 4.3.1 : 1797 ; 2701
UnixBench (2)
System Benchmarks Index Score x1 479.3
System Benchmarks Index Score x2 -
1.5- VPS 9 (Buffalo)
Unpublished__data
Code: (Select All)
Processor : Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
CPU Cores : 2 @ 2499.998 MHz
Kernel : 4.18.0-240.1.1.el8_3.x86_64
Scores for GeekBench 5.1.1, 5.1.0 and 4.3.1, in single and multi-core tests:
>> GeekBench 5.1.1 : 367 ; 646 (https://browser.geekbench.com/v5/cpu/6320829)
>> GeekBench 5.1.0 : 396 ; 654 (https://browser.geekbench.com/v5/cpu/6321825)
>> GeekBench 4.3.1 : 1931 ; 2995 (https://browser.geekbench.com/v4/cpu/16028249)
UnixBench (2)
System Benchmarks Index Score x1 436.0
System Benchmarks Index Score x2 837.7
2- Libvirt CPU Models :
Libvirt supports three ways to configure guests CPU models:
- Host passthrough - In this mode, the host CPU model, stepping, and features are faithfully passed to the guest while certain CPU features will still be filtered out by the hypervisor. This guest CPU model is the recommended CPU to use when live migration isn't needed.
This mode was used for VPS-9 @Phoenix, @Buffalo and @Seattle.
- Named models - This is a set of predefined named CPU models supported by QEMU and that correspond to specific generations of CPUs released by hardware vendors. These named CPUs are typically used when live migration between hosts with differing hardware is mission critical.
This mode was used for VPS-9 @Atlanta.
- Host model - This is the default mode of libvirt that tries to use QEMU's "Named models" to automatically chose a CPU model that is as close to the host CPU as possible while adding any extra flag optimising both host CPU matching and VM live migration.
This mode was used for VPS-9 @L.A.
Armed with the above information in mind, we're now in a position to fine-tune our virt-install command even further, like so:
Code: (Select All)
virt-install --virt-type=kvm --hvm --arch=x86_64 --machine=pc \
--name=centos8Phoenix \
--os-type=linux --os-variant=centos8 \
--ram=8192 --vcpus=2 --cpu host-passthrough \
--disk path=/media/phoenix.img \
--import
Again, why is it important to add the cpu flag in the command?.. Answer, to avoid ending up with a host-model CPU (on my system, it's Model: IvyBridge-IBRS that resolve to : Intel Xeon E3-12xx v2 inside the guest) instead of the more powerful host-passthrough model that will resolve to your own host CPU (ie Intel® Core i5-3470 CPU @ 3.20GHz on my KVM system.)
In the next post we'll talk a bit about disk I/O fine-tuning.