- 28 May 2009
RTT Delays caused by a large TCP flow directed at a home gateway
Methodology: There are two nodes, one is a home gateway and the other is the BITS server at UIC. The gateway (running Openwrt) has available 3 different pings. ICMP (standard ping utility) and two programs that we wrote that use TCP or UDP. We wrote these utilities because we were seeing strange behavior when just using the standard ICMP ping.
The gateway begins pinging the BITS server (conducting experiments one protocol at a time). We then start a TCP flow (using the nuttcp utility) from BITS directed at the home gateway. We gradually increase that flow and measure the RTT times of the pings. Below are some results:
Ingress traffic and delays
Methodology: We have a linksys router with an ingress rate limit of 200kbit/s. We then have an PicoStation2 wireless AP that uses the Linksys as a gateway (using Ethernet). We then conduct experiments by running iperf on a computer outside our network (as the client) sending traffic to a node within our network. We then send pings from the same node within our network and monitor the effect of the incoming traffic on ping RTT. Also, we will gradually rate limit the PicoStation 's ingress to see how that will effect RTT.