Is is possible to stress test srsRAN DU Without UE Connection

[Zplusless]

Hi everyone! I'm currently working on a project involving srsRAN. My goal is to conduct stress testing on the Distributed Unit (DU) component of srsRAN. My question is: Is it possible to simulate the network load or stress conditions to evaluate the DU's performance/computation resource usage under different workloads without attaching a UE? I see some code in apps/examples/. Is it possible to run DU individually and simulate UL/DL loads?

Thanks in advance!

[pgawlowicz]

Hi,
Please check the test_mode in the gnb config: https://docs.srsran.com/projects/project/en/latest/user_manuals/source/config_ref.html
and also see answere here: #446

[pgawlowicz]

you can set the cqi for the test_ue.

also remove this part, as this fixes the MCS:

cell_cfg:
...
  pdsch: 
    min_ue_mcs: 10                  # Optional UINT (0). Sets a minimum PDSCH MCS value to be used for all UEs. Supported: [0 - 28].
    max_ue_mcs: 10                  # Optional UINT (28). Sets a maximum PDSCH MCS value to be used for all UEs. Supported: [0 - 28].

  pusch: 
    min_ue_mcs: 10                  # Optional UINT (0). Sets a minimum PUSCH MCS value to be used for all UEs. Supported: [0 - 28].
    max_ue_mcs: 10                  # Optional UINT (28). Sets a maximum PUSCH MCS value to be used for all UEs. Supported: [0 - 28].

[Zplusless]

Thanks！

[Zplusless]

Hi, just noticed that the master branch is updated and an RU emulator application is added. Is it possible to generate dynamic bit streams to test gNB (for example, replaying a 5G PUSCH/PDSCH data trace) with this emulator?

[sauka]

Hi @Zplusless ,

We've added a short description of the RU emulator application added in the recent release, please find it here #517

[Zplusless]

Hi @sauka, thanks for your reply. After a brief review of ru_emulator.cpp, it seems that it is possible to generate dynamic workloads for gNB.

[Zplusless]

Hi @sauka , Thanks for the super cool RU emulator mode. We have learned how UL streams are generated in ru_emulator.cpp. Could you please provide some guidance on how the DL streams are generated on the gNB side in RU emulator mode?

Thanks in advance!

[sauka]

Hi @Zplusless

It doesn't matter whether a real RU or the RU emulator is used in the test, the gNB will generate DL traffic according to the cell and slot configuration (e.g. TDD pattern) and in the test mode it will try to push as much traffic as possible (e.g. it might be limited by the host system performance).

Regarding your other question about dynamic workloads: the RU emulator doesn't generate IQ samples based on what PDSCH carries, it replies with random data based on what the DU commands it

[Zplusless]

Hi @sauka，thanks for your reply. We are trying to modify the test mode so that it could generate on demand (with a data trace as input) instead of as much traffic as possible. This could help to emulate real-world scenarios for research, considering srsUE only support 20MHz and Amarisoft UE could not generate large enough throughput.

It doesn't matter whether a real RU or the RU emulator is used in the test

We have tried test mode both with an srsUE or with an ru_emulator, we understand that the DL traffic is generated by filling meaningless bits at gNB. If I'm right, the role of an srsUE or an ru_emulator is to make gNB to start working.

the gNB will generate DL traffic according to the cell and slot configuration (e.g. TDD pattern) and in the test mode it will try to push as much traffic as possible (e.g. it might be limited by the host system performance).

We have checked files such as mac_test_mode_adapter.cpp and mac_cell_processor.cpp and found some key functions for data processing such as assemble_dl_data_request(). However, we are still confusing and find it difficulty in looking for the very source of DL data generation in test mode. We are assuming that only phy and mac layer work in test mode, so the data source may work above MAC. But we are still struggling finding where random bits are generates and passed to MAC since upper layers are not working.

Regarding your other question about dynamic workloads: the RU emulator doesn't generate IQ samples based on what PDSCH carries,

Understand. The RU will drop DL UP data and print CP info in the console.

it replies with random data based on what the DU commands it

Do you mean the UL grants that the DU MAC generates? So the UL is also in "as much as possible" mode?