Have a VPLS network connecting 6 different sites back to our data center. Each site has a dedicated 50 MB connection with bursting up to 100 MB. Trying to do a file migration from one site to the data center and I'm only getting .5-.7 MBytes the other sites are fine. Tried doing some captures on the VPLS interface while testing with iperf and I'm seeing retransmissions and unseen segment packets. Not sure if this could be causing the slowness or not. Here is a snippet of the capture https://www.cloudshark.org/captures/619976c41207 asked 13 Apr '16, 21:32  wargresh edited 14 Apr '16, 05:47 |
This is a static archive of our old Q&A Site.
Please post any new questions and answers at ask.wireshark.org.
you have not taken capture from beginning but it looks that RTT is around 2 ms,regarding your IPERF Testing on port 5001,i could see your iperf client only sending 17032 bytes/rtt so 17032*8 = 136256 bits and RTT is approx 2 ms = 136256/.002 = 68 mbits/sec.In your capture there are retransmission as well with RTO of 300 ms,no delayed ack.so the slowness could be due to 10.10.3.24 stop acking packets that caused RTO to kick in.if you can take a capture on 10.10.3.24 you could find out the reason.
or i feel 10.10.3.24 didn't received the packets,could be packet dropped.
Kishan,
Thanks for the information. I have performed a capture on 10.10.3.24 specifying the iperf traffic. Here is the link: https://www.cloudshark.org/captures/2229954b2e30
yes,there is packet drop,10.10.3.24 didnt received packets and because iperf is only sending 16k bytes/rtt and waits for ack before sending any data,due to this behaviour fast retransmit is not triggered and sender is waiting for rto