Why I am not getting max enq_qdepth of above 60 on the 2nd experiment on FT?

[Nagmat Nazarov]

Dear Community

I am doing experiment on Fabric testbed. My slice id is : fdb44db5-abfb-406d-843b-0dd3d18c99a5

The architecture is attached as lab1.png.

I am sending probing packet from server1 to server2 and measuring enq_qdepth status every 0.001 seconds.

On the probing packet number 5 and 600 I am initializing the congestion process as : “iperf3 -c 192.168.2.10 -u -b -1 -t 20″

From the first congestion I am getting the expected results but I am not getting expected result from the second congestion as shown in the figure temp.png.

Even if I see that sender is sending 14G of udp packets and receiver is receiving only 995M only(since I am limiting the bandwidth with tc command : ”

`sudo tc qdisc add dev enp7s0 root handle 1:0 netem delay 1ms
sudo tc qdisc add dev enp7s0 parent 1:1 handle 10: tbf rate 1gbit buffer 16000000 limit 30000000`")

Why I am not getting max enq_qdepth of above 60 on the experiment? May it be related to Fabric architecture?

Kind regards,

Nagmat

• This topic was modified 1 year, 1 month ago by Nagmat Nazarov.

[Ilya Baldin]

It is quite likely that this is related to either switch behavior or the behaviour of SharedNICs. You appear to be using SharedNICs, and for most control over your experiment you should be using SmartNICs as those are not shared with anyone.

[Author]

Posts