Add support for uperf driver

venkataanil commented 10 months ago

Similar to iperf, user can run uperf along with netperf using "--uperf" option

benchmark-wrapper is used a reference for parsing 1) user options and creating uperf config file (input to uperf client command) 2) uperf output

uperf driver supports only TCP_STREAM, UPD_STREAM, TCP_RR, and UDP_RR tests. For each test in full-run.yaml, Uperf driver will create a uperf profile file inside the client pod and uses it to run the test. Parallelism is implemented using uperf's nproc option

uperf server can't be run using "&&" option in the same container inside server pod i.e "netserver && iperf3 -s -p 22865 && uperf -s -v -P 30000 && sleep 10000000" So we are creating a separate container in server pod for uperf server.

Type of change

[ ] Refactor
[x] New feature
[ ] Bug fix
[ ] Optimization
[ ] Documentation Update

Description

Related Tickets & Documents

Related Issue #
Closes #

Checklist before requesting a review

[ ] I have performed a self-review of my code.
[ ] If it is a core feature, I have added thorough tests.

Testing

Please describe the System Under Test.
Please provide detailed steps to perform tests related to this code change.
How were the fix/results from this change verified? Please provide relevant screenshots or results.

SachinNinganure commented 9 months ago

Hello @venkataanil Can we also have the driver details on RR output

venkataanil commented 9 months ago

Hello @venkataanil Can we also have the driver details on RR output

done, thanks

rsevilla87 commented 9 months ago

"netserver && iperf3 -s -p 22865 && uperf -s -v -P 30000 && sleep 10000000" So we are creating a separate container in server pod for uperf server.

Why?, because iperf and uperf server mode run in foreground?

You can always use & to run them in background with something like:

$ (netserver & iperf3 -s -p 22865 & uperf -s -v -P 30000) && sleep inf

$ ps -ef | egrep "uperf|netserver|iperf"
egrep: warning: egrep is obsolescent; using grep -E
rsevilla  839931  839929  0 22:57 pts/3    00:00:00 iperf3 -s -p 22865
rsevilla  839932  839929  0 22:57 pts/3    00:00:00 uperf -s -v -P 30000
rsevilla  839933    3241  0 22:57 ?        00:00:00 netserver

venkataanil commented 9 months ago

Thanks, I tried only in foreground . Moving them to containers is a good option. Lets continue it.

rsevilla87 commented 9 months ago

Thanks, I tried only in foreground . Moving them to containers is a good option. Lets continue it.

I'm not sure what changes you've added in the last commit but I still see multiple containers for the servers

jtaleric commented 9 months ago

Thanks, I tried only in foreground . Moving them to containers is a good option. Lets continue it.

I'm not sure what changes you've added in the last commit but I still see multiple containers for the servers

@rsevilla87 I would like to get 1 more +1 -- let me know if you feel there is anything else for @venkataanil to address.

jtaleric commented 9 months ago

Does uperf not provide re-transmission info?

+---------------------+---------+------------+-------------+--------------+---------+--------------+-----------+----------+---------+------------+
|        TYPE         | DRIVER  |  SCENARIO  | PARALLELISM | HOST NETWORK | SERVICE | MESSAGE SIZE | SAME NODE | DURATION | SAMPLES | AVG VALUE  |
+---------------------+---------+------------+-------------+--------------+---------+--------------+-----------+----------+---------+------------+
| TCP Retransmissions | netperf | TCP_STREAM | 1           | false        | false   | 1024         | false     | 30       | 3       | 38.333333  |
| TCP Retransmissions | uperf   | TCP_STREAM | 1           | false        | false   | 1024         | false     | 30       | 3       | 0.000000   |
| TCP Retransmissions | netperf | TCP_STREAM | 2           | false        | false   | 1024         | false     | 30       | 3       | 47.000000  |
| TCP Retransmissions | uperf   | TCP_STREAM | 2           | false        | false   | 1024         | false     | 30       | 3       | 0.000000   |
| TCP Retransmissions | netperf | TCP_STREAM | 1           | false        | false   | 8192         | false     | 30       | 3       | 40.333333  |
| TCP Retransmissions | uperf   | TCP_STREAM | 1           | false        | false   | 8192         | false     | 30       | 3       | 0.000000   |
| TCP Retransmissions | netperf | TCP_STREAM | 2           | false        | false   | 8192         | false     | 30       | 3       | 170.000000 |
| TCP Retransmissions | uperf   | TCP_STREAM | 2           | false        | false   | 8192         | false     | 30       | 3       | 0.000000   |
| UDP Loss Percent    | netperf | UDP_STREAM | 1           | false        | false   | 1024         | false     | 30       | 3       | 2.435977   |
| UDP Loss Percent    | uperf   | UDP_STREAM | 1           | false        | false   | 1024         | false     | 30       | 3       | 0.000000   |
+---------------------+---------+------------+-------------+--------------+---------+--------------+-----------+----------+---------+------------+

venkataanil commented 9 months ago

Thanks, I tried only in foreground . Moving them to containers is a good option. Lets continue it.

I'm not sure what changes you've added in the last commit but I still see multiple containers for the servers

Lets go with multiple containers approach only. So I retained multiple containers code in this patch.

venkataanil commented 9 months ago

Does uperf not provide re-transmission info?

+---------------------+---------+------------+-------------+--------------+---------+--------------+-----------+----------+---------+------------+
|        TYPE         | DRIVER  |  SCENARIO  | PARALLELISM | HOST NETWORK | SERVICE | MESSAGE SIZE | SAME NODE | DURATION | SAMPLES | AVG VALUE  |
+---------------------+---------+------------+-------------+--------------+---------+--------------+-----------+----------+---------+------------+
| TCP Retransmissions | netperf | TCP_STREAM | 1           | false        | false   | 1024         | false     | 30       | 3       | 38.333333  |
| TCP Retransmissions | uperf   | TCP_STREAM | 1           | false        | false   | 1024         | false     | 30       | 3       | 0.000000   |
| TCP Retransmissions | netperf | TCP_STREAM | 2           | false        | false   | 1024         | false     | 30       | 3       | 47.000000  |
| TCP Retransmissions | uperf   | TCP_STREAM | 2           | false        | false   | 1024         | false     | 30       | 3       | 0.000000   |
| TCP Retransmissions | netperf | TCP_STREAM | 1           | false        | false   | 8192         | false     | 30       | 3       | 40.333333  |
| TCP Retransmissions | uperf   | TCP_STREAM | 1           | false        | false   | 8192         | false     | 30       | 3       | 0.000000   |
| TCP Retransmissions | netperf | TCP_STREAM | 2           | false        | false   | 8192         | false     | 30       | 3       | 170.000000 |
| TCP Retransmissions | uperf   | TCP_STREAM | 2           | false        | false   | 8192         | false     | 30       | 3       | 0.000000   |
| UDP Loss Percent    | netperf | UDP_STREAM | 1           | false        | false   | 1024         | false     | 30       | 3       | 2.435977   |
| UDP Loss Percent    | uperf   | UDP_STREAM | 1           | false        | false   | 1024         | false     | 30       | 3       | 0.000000   |
+---------------------+---------+------------+-------------+--------------+---------+--------------+-----------+----------+---------+------------+

I used benchmark-wrapper and benchmark-operator as base for running uperf and parsing results. uperf wrapper https://github.com/cloud-bulldozer/benchmark-wrapper/blob/master/snafu/benchmarks/uperf/uperf.py was not having info related to re-transimmion, so I couldn't add it. I will push another PR for re-transmition and additional changes.

cloud-bulldozer / k8s-netperf