This is a static archive of our old Q&A Site. Please post any new questions and answers at ask.wireshark.org.

Strange RTT graph

0

Hi!

What is the reason of this kind of RTT graph?

BR.alt text

asked 26 Nov '12, 00:13

Nurik's gravatar image

Nurik
1111
accept rate: 0%

can you please add a "Throughput Graph" and a "Time-Sequence Graph (tcptrace)".

(26 Nov '12, 02:33) Kurt Knochner ♦

Throuput graphalt text

(04 Dec '12, 04:23) Nurik

Time-Sequence Graph (tcptrace)alt text

(04 Dec '12, 04:24) Nurik

2 Answers:

1

The vertical "lines" in your RTT graph might be wrongly assumed high delay packets, when wireshark is checking a certain sequence number for the time it takes TCP to ACK the data. When you have packet loss, this can be a typical graphical outcome of wireshark interpreting constant ACK numbers while waiting for a retransmission of lost data.

That could in theory also explain why your UDP transfer gets higher throughput, if the UDP application doesn't care for packet loss, while TCP very well does - check for tcp.analysis.flags

answered 26 Nov '12, 02:06

Landi's gravatar image

Landi
2.3k51442
accept rate: 28%

0

Looks pretty normal to me... you've got tons of times below 0.05 seconds, and sometimes it goes up when there is contention, which is caused by buffering delays.

answered 26 Nov '12, 00:22

Jasper's gravatar image

Jasper ♦♦
23.8k551284
accept rate: 18%

Thank you, but how it can be solved? Previous RTT graph was made on client side logs. I think i have low throughput due to this delays. Below RTT graph of same session from point closer to server. alt text

(26 Nov '12, 00:35) Nurik

and what it means "which is caused by buffering delays." ?

BR

(26 Nov '12, 00:36) Nurik

Additional info regarding this issue, with UDP I'm getting 30Mbps, but with TCP no more then 10Mbps.

(26 Nov '12, 00:38) Nurik