search

Gradiant / 5g-charts

Helm charts for 5G Technologies
Apache License 2.0
108 stars 42 forks source link

PacketRusher Helm with several UEs and GNB parameters #165

Closed rodrigosanzsanz closed 4 months ago

rodrigosanzsanz commented 5 months ago

Components and subsystems

PacketRusher Image

Proposed functionality

Hello!

Great work! We are facing limitations to create a big number of ues and inject traffic to Open5gs. Maybe PacketRusher is a better option to simulate real traffic, but in the Helm of Packet rusher only 1 UE is launched in the gNB. Is it possible to improve this parameter config? Similar to UERANSIM.. Perhaps there is another limitations or the performance with big amount of UEs have some problem.

Could you provide some info?

Thanks!

External dependencies

PacketRusher

avrodriguezgrad commented 5 months ago

Hi @rodrigosanzsanz

PacketRusher chart (and image also) was created to work only with the simple configuration of the tool. This was made in that way to check its behaviour as a container and as a pod.

Roadmap contemplates to allow more configurations, but it will take us some time to implement it. If you want, you can try to change the image and/or the chart, and create a PR.

BR, Álvaro

avrodriguezgrad commented 4 months ago

Hi @rodrigosanzsanz

Chart and image were changed to support the mode selection of PacketRusher. Using args in values.yaml, you can use ue, multi-ue etc, and put the parameters you want.

FYI, if you want to have multiple UEs and support data plane traffic, I have to tell you that that is impossible right now. I forward you to this issue for more context: https://github.com/HewlettPackard/PacketRusher/issues/88

BR, Álvaro

rodrigosanzsanz commented 4 months ago

Thanks!! I see the updates and the issu with multiUE. Really appreciate your info!Thanks

linouxis9 commented 4 months ago

Thanks a lot for your work @avrodriguezgrad :-) To be noted, the --dedicatedGnb option could probably made to work in k8s using Multus and creating dedicated interfaces in the pod for each gNodeB, but it won't scale to many gNodeBs.