bw (bandwidth) config in p4app.json does not take effect about p4-utils HOT 8 CLOSED

Are you sure your traffic is crossing that link? I believe it is not even crossing the link. Getting 20Gbps through bmv2 switch is very unlikely.

Try to check where traffic goes first, I think the problem resides there.

from p4-utils.

I run the example in p4-utils, the iperf test is the same (around 20Gbps)

mininet> iperf s1 s2
*** Iperf: testing TCP bandwidth between s1 and s2 
[            Level 25] newfn: *** Iperf: testing TCP bandwidth between s1 and s2 
 
*** Results: ['18.3 Gbits/sec', '18.3 Gbits/sec']
[            Level 25] newfn: *** Results: ['18.3 Gbits/sec', '18.3 Gbits/sec']

from p4-utils.

I believe that iperf test is not using the interfaces at all.

Can you do the iperf from hosts?

from p4-utils.

Iperf from hosts perform as expected, around 20Mbps, which is as same as config.

If iperf test on switch is not using interface, it test on what?

from p4-utils.

No it is not, switches are located at the main namespace, if you assign ips to them (which i do not recommend, and believe it is explained somewhere) those ips get configured at the main namespace, meaning that when you send traffic from SX to SY they are directly forwarded locally without using any link.

if you do route -n you will see that both routes are local at the main namespace.

Hope this helps, if not I can try to explain it with further detail.

from p4-utils.

As traffic between switches is directly forwarded locally, Is there some other way to restrict the bandwidth between switches traffic?

from p4-utils.

There are two possible options I believe:

• Make switches get their own network namespace, but then many p4-utils things that assume that switches can be configured/accesses locally might not work.

• Just attach some nodes to those switches and do the communication from them (it is not the same but close..).

from p4-utils.

if you do route -n you will see that both routes are local at the main namespace.

Thank you for your explanation, but I did route -n but cannot find the printout is related to the mininet ips. (33 is VM intranet IP)

$ route -n
Kernel IP routing table
Destination     Gateway         Genmask         Flags Metric Ref    Use Iface
0.0.0.0         33.33.33.21     0.0.0.0         UG    100    0        0 ens3
33.33.33.0      0.0.0.0         255.255.255.0   U     0      0        0 ens3
33.33.33.21     0.0.0.0         255.255.255.255 UH    100    0        0 ens3

There is some mininet info. (still the p4-utils example)

s1 -> Thrift port: 9090
s2 -> Thrift port: 9091
s3 -> Thrift port: 9092
**********
Network configuration for: h1
Default interface: h1-eth0      10.0.1.1        00:00:0a:00:01:01
**********
**********
Network configuration for: h2
Default interface: h2-eth0      10.0.2.2        00:00:0a:00:02:02
**********
**********
Network configuration for: h3
Default interface: h3-eth0      10.0.3.3        00:00:0a:00:03:03
**********
**********
Network configuration for: h4
Default interface: h4-eth0      10.0.3.4        00:00:0a:00:03:04
**********

from p4-utils.