QUESTION: Background traffic generator on gNodeB

[jaivra]

Hello, I'm having some problems with background traffic generation on base station. I tried your example (simulations/NR/bgTraffic) in two different scenarios, one with 10 background UEs each generates 1000B packet every 10ms, while the other with 200 background UEs each generates 1000MB packet every 10ms. Observing the results I noticed that the uplink delay (statistic cbrFrameDelay:mean in SingleCell_Standalone.server.app[0]) remains substantially the same.

I wonder if this is the right procedure to generate some background traffic at the base station and if the generator uses the same resource blocks of the MAC scheduler as for the UE.

Thanks for your help, Valerio.

[giovanninardini]

Hello,

my guess is that the UE is very close to the gNB (50m), hence its CQI is always very high compared to the the CQI of background UEs. As a result, the MaxC/I scheduler (the default one) mostly prioritizes that UE over the background UEs, hence the cbrFrameDelay does not change much when increasing the background traffic.
Please try to change the position of the UE (you should find the variable $dist somewhere in the omnetpp.ini file) and see what happens.

Best regards. Giovanni

[jaivra]

Hello, Thanks for the answer. So, I followed your suggestions, and changed your example (simulations/NR/bgTraffic), moving the UE away from the gnb causes the delay to increase. I tried three different scenarios with numBgCells = 0 and varying the numBgUEs (0, 10, 20). I investigated the uplink delay (statistic cbrFrameDelay:mean in SingleCell Standalone.server.app[0]) and obtained the following results:

• numBgUEs = 0 -> cbrFrameDelay:mean (scalar) = 0.026326
• numBgUEs = 10 -> cbrFrameDelay:mean (scalar) = 0.232496
• numBgUEs = 20 -> cbrFrameDelay:mean (scalar) = 0.034010

So, the delay increases significantly from 0 to 10 BgUEs, but at 20 BgUEs, the delay is comparable to the scenario with no background traffic. How is this possible? What am I doing incorrectly?

PS: I've attached the configuration file, so you can reproduce the results.

Thanks for your help, Best regards, Valerio. omnetpp.txt

[christianquadri]

Hello, I have investigated the question of Valerio. I found that the background UEs are considered only if MAXCI scheduling discipline is used, while Valerio and I used PF and DRR last year.

Second, it seems that the resource blocks are firstly allocated to the foreground UEs and the remaining to the background ones. Is that correct? I did some tests with MAXCI and forcing bgUE CQI fixed to 15 UL&DL and I saw that the resource blocks of gNB are completely used, but the saturation of the gNB does not affect the foreground UEs.

Resource block usage (background UEs start at 30 s)

RTT (unaffected by saturation)

Thank you for your time

Christian