[Xen-API] Low CPU benchmark performance on PV VM

["Jeroen van Kessel" <jkessel@xxxxxx>]

Hi Xen-community,

I am running a CPU benchmark tool (LINPACK) on a PV VM inside Xen 4.5.1 on
Ubuntu 15.10 x64 on a IBM x3550 M4 server. This tool should consume all
possible CPU cycles available. I allocate 4 vCPUs by defining this in the
Xen PV (test.cfg). However, only 1 core and 4 threads are used:

CPU frequency:    2.494 GHz
Number of CPUs: 1
Number of cores: 1
Number of threads: 4

This is what lscpu says inside this Xen PV VM:

    Architecture:          x86_64
    CPU op-mode(s):        32-bit, 64-bit
    Byte Order:            Little Endian
    CPU(s):                4
    On-line CPU(s) list:   0-3
    Thread(s) per core:    4
    Core(s) per socket:    1
    Socket(s):             1
    NUMA node(s):          1
    Vendor ID:             GenuineIntel
    CPU family:            6
    Model:                 62
    Model name:            Intel(R) Xeon(R) CPU E5-2609 v2 @ 2.50GHz
    Stepping:              4
    CPU MHz:               2500.062
    BogoMIPS:              5000.12
    Hypervisor vendor:     Xen
    Virtualization type:   full
    L1d cache:             32K
    L1i cache:             32K
    L2 cache:              256K
    L3 cache:              10240K
    NUMA node0 CPU(s):     0-3

A HVM VM DO get all possible cores allocated inside the virtual node (see
below). These nodes have significant better performance then the Xen PV
virtual node.

    CPU frequency:    2.494 GHz
    Number of CPUs: 1
    Number of cores: 4
    Number of threads: 4

This is the DomU Xen host machine lscpu:

    lscpu

    Architecture:          x86_64
    CPU op-mode(s):        32-bit, 64-bit
    Byte Order:            Little Endian
    CPU(s):                8
    On-line CPU(s) list:   0-7
    Thread(s) per core:    1
    Core(s) per socket:    4
    Socket(s):             2
    NUMA node(s):          1
    Vendor ID:             GenuineIntel
    CPU family:            6
    Model:                 62
    Model name:            Intel(R) Xeon(R) CPU E5-2609 v2 @ 2.50GHz
    Stepping:              4
    CPU MHz:               2500.062
    BogoMIPS:              5000.12
    Hypervisor vendor:     Xen
    Virtualization type:   none
    L1d cache:             32K
    L1i cache:             32K
    L2 cache:              256K
    L3 cache:              10240K
    NUMA node0 CPU(s):     0-7


Xen vCPU list:

xl vcpu-list

Domain-0 0 0 7 -b- 10.1 all / all
Domain-0 0 1 2 -b- 6.5 all / all
Domain-0 0 2 4 r-- 2.6 all / all
Domain-0 0 3 0 r-- 3.9 all / all
Domain-0 0 4 3 -b- 4.4 all / all
Domain-0 0 5 6 -b- 2.6 all / all
Domain-0 0 6 5 -b- 4.7 all / all
Domain-0 0 7 7 -b- 2.9 all / all
test 3 0 1 -b- 1.5 0-3 / all
test 3 1 0 -b- 1.8 0-3 / all
test 3 2 0 -b- 0.7 0-3 / all
test 3 3 2 -b- 0.6 0-3 / all

xen DomU PV VM config:

    cat test.cfg

    bootloader = '/usr/lib/xen-4.5/bin/pygrub'

    vcpus       =  '4'
    memory      =  '2048'
    cpus        = "0-3"


Is there any options to give the paravirtualized guest the host-cpu
topology? In other words, how do I get Xen to use more vCPU cores?

Thanks!


_______________________________________________
Xen-api mailing list
Xen-api@xxxxxxxxxxxxx
http://lists.xen.org/cgi-bin/mailman/listinfo/xen-api