Open Jason-lmx opened 1 year ago
To do this, make sure to use congestion control that supports ECT(1) traffic on the sender (e.g., TCP-Prague) and set the proper tcp_ecn parameter on both sender and receiver. Taking TCP-Prague as an example,
On the sender side, configure Prague congestion control and tcp_ecn:
sysctl -w net.ipv4.tcp_congestion_control=Prague
sysctl -w net.ipv4.tcp_ecn=3
On the client side, configure tcp_ecn:
sysctl -w net.ipv4.tcp_ecn=3
If above configuration is not doable, make sure you already load the module.
I am very interest in testing L4S in my network. I am trying to understand how it affects the terminals. "sysctl" seems to be a global behavior. Does TCP prague support customizing the parameters of a single flow,except ‘ecn’ ?
Also, I would like to know how QUIC supports BBRv2 and L4S, any suggestions are welcome.
Indeed, sysctl is a global behaviour. But if you run your L4S in the network namespace it will not affect the root namespace. Not sure about your meaning for "customising the parameters of a single flow", but if your goal is to test L4S without modifying all your current setup, the simplest way is to run the application using namespace or even use dedicated machines with Prague kernel.
Related to QUIC support L4S, AFAIK from the thread (https://developer.apple.com/forums/thread/720486), the QUIC can support L4S under Developer setting. But related QUIC support BBRv2, I am not aware on that.
I mean could I set some parameters of TCP prague in SOCKET, such as rtt_target. Also,May I ask when a congestion event is detected, how does the sender adjust the congestion window? What is the difference between ‘prague_enter_loss’ and 'prague_enter_cwr'?Why did you use ‘prague_classic_ecn_fallback’ to update alpha?
modinfo tcp_prague systool -vm tcp_prague
I want to set a flow to ECT(1) as low-latency traffic. So I would like to know if I can selectively use TCP Prague as a congestion control method in socket programming, what parameters can be set, and how to set them. Looking forward to your reply.