1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
|
.. This work is licensed under a Creative Commons Attribution 4.0 International License.
.. http://creativecommons.org/licenses/by/4.0
=============================================
Low Latency Feature Configuration Description
=============================================
Introduction
------------
In KVM4NFV project, we focus on the KVM hypervisor to enhance it for NFV, by
looking at the following areas initially
* Minimal Interrupt latency variation for data plane VNFs:
* Minimal Timing Variation for Timing correctness of real-time VNFs
* Minimal packet latency variation for data-plane VNFs
* Inter-VM communication,
* Fast live migration
Configuration of Cyclictest
---------------------------
Cyclictest measures Latency of response to a stimulus. Achieving low latency
with the KVM4NFV project requires setting up a special test environment.
This environment includes the BIOS settings, kernel configuration, kernel
parameters and the run-time environment.
* For more information regarding the test environment, please visit
https://wiki.opnfv.org/display/kvm/KVM4NFV+Test++Environment
https://wiki.opnfv.org/display/kvm/Nfv-kvm-tuning
Pre-configuration activities
~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Intel POD10 is currently used as OPNFV-KVM4NFV test environment. The rpm
packages from the latest build are downloaded onto Intel-Pod10 jump server
from artifact repository. Yardstick running in a ubuntu docker container
on Intel Pod10-jump server will configure the host(intel pod10 node1/node2
based on job type), the guest and triggers the cyclictest on the guest using
below sample yaml file.
.. code:: bash
For IDLE-IDLE test,
host_setup_seqs:
- "host-setup0.sh"
- "reboot"
- "host-setup1.sh"
- "host-run-qemu.sh"
guest_setup_seqs:
- "guest-setup0.sh"
- "reboot"
- "guest-setup1.sh"
.. figure:: images/idle-idle-test.png
:name: idle-idle-test
:width: 100%
:align: center
.. code:: bash
For [CPU/Memory/IO]Stress-IDLE tests,
host_setup_seqs:
- "host-setup0.sh"
- "reboot"
- "host-setup1.sh"
- "stress_daily.sh" [cpustress/memory/io]
- "host-run-qemu.sh"
guest_setup_seqs:
- "guest-setup0.sh"
- "reboot"
- "guest-setup1.sh"
.. figure:: images/stress-idle-test.png
:name: stress-idle-test
:width: 100%
:align: center
The following scripts are used for configuring host and guest to create a
special test environment and achieve low latency.
Note: host-setup0.sh, host-setup1.sh and host-run-qemu.sh are run on the host,
followed by guest-setup0.sh and guest-setup1.sh scripts on the guest VM.
**host-setup0.sh**: Running this script will install the latest kernel rpm
on host and will make necessary changes as following to create special test
environment.
* Isolates CPUs from the general scheduler
* Stops timer ticks on isolated CPUs whenever possible
* Stops RCU callbacks on isolated CPUs
* Enables intel iommu driver and disables DMA translation for devices
* Sets HugeTLB pages to 1GB
* Disables machine check
* Disables clocksource verification at runtime
**host-setup1.sh**: Running this script will make the following test
environment changes.
* Disabling watchdogs to reduce overhead
* Disabling RT throttling
* Reroute interrupts bound to isolated CPUs to CPU 0
* Change the iptable so that we can ssh to the guest remotely
**stress_daily.sh**: Scripts gets triggered only for stress-idle tests. Running this script
make the following environment changes.
* Triggers stress_script.sh, which runs the stress command with necessary options
* CPU,Memory or IO stress can be applied based on the test type
* Applying stress only on the Host is handled in D-Release
* For Idle-Idle test the stress script is not triggered
* Stress is applied only on the free cores to prevent load on qemu process
**Note:**
- On Numa Node 1: 22,23 cores are allocated for QEMU process
- 24-43 are used for applying stress
**host-run-qemu.sh**: Running this script will launch a guest vm on the host.
Note: download guest disk image from artifactory.
**guest-setup0.sh**: Running this scrcipt on the guest vm will install the
latest build kernel rpm, cyclictest and make the following configuration on
guest vm.
* Isolates CPUs from the general scheduler
* Stops timer ticks on isolated CPUs whenever possible
* Uses polling idle loop to improve performance
* Disables clocksource verification at runtime
**guest-setup1.sh**: Running this script on guest vm will do the following
configurations.
* Disable watchdogs to reduce overhead
* Routes device interrupts to non-RT CPU
* Disables RT throttling
Hardware configuration
~~~~~~~~~~~~~~~~~~~~~~
Currently Intel POD10 is used as test environment for kvm4nfv to execute
cyclictest. As part of this test environment Intel pod10-jump is configured as
jenkins slave and all the latest build artifacts are downloaded on to it.
* For more information regarding hardware configuration, please visit
https://wiki.opnfv.org/display/pharos/Intel+Pod10
https://build.opnfv.org/ci/computer/intel-pod10/
http://artifacts.opnfv.org/octopus/brahmaputra/docs/octopus_docs/opnfv-jenkins-slave-connection.html
|
