Dear all, In the Linux kernel there are two variables that specify buffer size of TCP sender and receiver.
For instance, in the case of receiver, this variable is tcp_rmen (/proc/sys/net/ipv4/tcp_rmem). It contains the following values: 4096 87380 4194304 which stand for the minimum, default and maximum receive buffer for each TCP socket, respectively (values in bytes). In this example, the maximum value allowed for each TCP socket is 4MB. I would like to know to it is possible to set this values on a ns-2 TCP-sink. I conducted a simulation with the following receiver (using TCP LINUX/Cubic as a sender): set sink [new Agent/TCPSink/Sack1/DelAck] $sink set ts_echo_rfc1323_ true $ns attach-agent $n4 $sink $ns connect $tcp $sink I'm asking because I'm simulating on 10Gbps speeds and packets are arriving out-of order, and I cannot see any discard on the receiver side (so it must be buffering the data until it receives the missing packet). Analyzing the trace, I calculated the amount of packets that arrived before the missing packet finally arrived, which was 550810. Each packet in my case has a size of 1500 bytes, so the receiver buffer stored 826.215 MB! I would like to limit this receiver buffer side to see how it would perform when there is no more buffer space left. Thanks for your attention, Giovane
