UP load test using multi-ue

[abousselmi]

Hello @linouxis9

Were you able to send traffic using more than 1 UE ?

If I create 2 UEs with multi-ue, for some reason yet unknown to me (gtp5g kernel module limitation ?) only 1 single UE VRF can be used to send traffic (more specifically the first one that is used). I was able to reach roughly 5 Gbps using iperf3.

In this configuration, we should use a separate gNB per UE,

I tried to use TAP instead of VRF but it seems there is a problem with the CLI, it doesn't take into account the --tunnel-vrf false into account. I did not yet got the time to look at it in more details.

Thanks is advance.

[linouxis9]

Hi @abousselmi,

Yes, of course. With something like: sudo ./packetrusher multi-ue -n 2 --tunnel --dedicatedGnb.
I tried up to like 10 UEs, others users were successful as well.
Make sure your second gNodeB works (eg. you can ping its IP from outside).

Could you share some pcap?

Thanks and cheers,
Valentin

[linouxis9]

Thanks for the fast reply! I am running on a docker container on top of a Kind Kubernetes cluster. I may indeed need Multus in this case. I'll try it out...

If you are not planning to spawn millions of UEe, you can always run one container per UE with each having its own gNodeB -> it's the best for dataplane performance testing :-)

[alekodu]

So I thought as well and it works if I create a separate k8s deployment per UE-gNB but it does not work if I use the same deployment and scale to a larger number of replicas. With the scale up option, communication between UEs-gNBs is stopped and the AMF pod is restarted:

Latest replica:

Older deployment pods:

It would be more friendly if the user could just scale up the replicas. Do you know why this happens?

[linouxis9]

Make sure you are using a different gnbid for each gNodeB and a distinct MSIN for each UE. If you do not do that, the first UE to be registered will be released by the Core Network when the second UE with the same ID (IMSI/MSIN) will be released/freed by the Core Network!
I guess reusing the same deployment make each of the created pod using the same ConfigMap and as such the same MSIN and gnbId!

[alekodu]

You are right! Thanks for the explanation... Very useful tool you are building!

[linouxis9]

You are right! Thanks for the explanation... Very useful tool you are building!

No problem! Happy to help 😁