diff options
author | Tibor Frank <tifrank@cisco.com> | 2019-02-05 10:20:41 +0100 |
---|---|---|
committer | Tibor Frank <tifrank@cisco.com> | 2019-02-05 13:30:25 +0000 |
commit | 124101d22151239b0411a73ae4d2bf8d70970937 (patch) | |
tree | 3910b6e04d4737cbfc3295a25f86e7aaa3050d6a /docs/report/introduction/methodology_kvm_vms_vhost_user.rst | |
parent | a221ffe6144eb0f372521fbbc828b8a225af12cd (diff) |
CSIT-1420: Split methodology section to more files
Change-Id: I861e578434abdf72244d684fca8cfd66e1db9c28
Signed-off-by: Tibor Frank <tifrank@cisco.com>
Diffstat (limited to 'docs/report/introduction/methodology_kvm_vms_vhost_user.rst')
-rw-r--r-- | docs/report/introduction/methodology_kvm_vms_vhost_user.rst | 23 |
1 files changed, 23 insertions, 0 deletions
diff --git a/docs/report/introduction/methodology_kvm_vms_vhost_user.rst b/docs/report/introduction/methodology_kvm_vms_vhost_user.rst new file mode 100644 index 0000000000..34e3bc0447 --- /dev/null +++ b/docs/report/introduction/methodology_kvm_vms_vhost_user.rst @@ -0,0 +1,23 @@ +KVM VMs vhost-user +------------------ + +FD.io CSIT performance lab is testing VPP vhost with KVM VMs using +following environment settings: + +- Tests with varying Qemu virtio queue (a.k.a. vring) sizes: [vr256] + default 256 descriptors, [vr1024] 1024 descriptors to optimize for + packet throughput. +- Tests with varying Linux :abbr:`CFS (Completely Fair Scheduler)` + settings: [cfs] default settings, [cfsrr1] CFS RoundRobin(1) policy + applied to all data plane threads handling test packet path including + all VPP worker threads and all Qemu testpmd poll-mode threads. +- Resulting test cases are all combinations with [vr256,vr1024] and + [cfs,cfsrr1] settings. +- Adjusted Linux kernel :abbr:`CFS (Completely Fair Scheduler)` + scheduler policy for data plane threads used in CSIT is documented in + `CSIT Performance Environment Tuning wiki + <https://wiki.fd.io/view/CSIT/csit-perf-env-tuning-ubuntu1604>`_. +- The purpose is to verify performance impact (MRR and NDR/PDR + throughput) and same test measurements repeatability, by making VPP + and VM data plane threads less susceptible to other Linux OS system + tasks hijacking CPU cores running those data plane threads. |