You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
log_performance: # boolean, use false to omit performance test details in logs/result files (only pass or fail)
112
120
```
113
121
114
122
A current version of Python is recommended to run the tests. As of writing the minimum version to avoid warnings would be 3.7. However, the tests have been successfully run up to version 3.11, the latest active release as of writing. The same is true of pip, which should be a current version (23.0 as of writing, but this should be upgraded in the following steps).
### Objective(s): A RFC2544 performance test to evaluate a system's (relative) performance when running DPDK workloads. This relies on a trafficgen client and containers built from [netgauge](https://github.com/redhat-eets/netgauge)
4
+
5
+
### Test procedure
6
+
7
+
* Reset the VFs
8
+
```
9
+
echo 0 > /sys/class/net/$PF/device/sriov_numvfs
10
+
```
11
+
12
+
* Ensure the reset succeeds (or check no VF exists under the ```$PF``` by ensuring ```sriov_numvfs``` is 0
13
+
14
+
* Ensure the trafficgen server ports are on the same numa node, repeat with the two dut server ports
15
+
```
16
+
cat /sys/bus/pci/devices/<pci_address>/numa_node
17
+
```
18
+
19
+
* Ensure the trafficgen server has 2 1GB hugepages, repeat with the dut server
20
+
21
+
* On the trafficgen, get 7 isolated CPUs from the numa node associated with the trafficgen ports, repeat on the dut server with 3 isolated CPUs
22
+
23
+
* On the dut server, create 1 VF on each PF, setting spoof checking off and trust mode on, and bind to vfio-pci
0 commit comments