Ask Your Question

Revision history [back]

stumbling over the use of io,stat,0,SUM

I don't understand the output I get when I try to sum tcp.time_delta in a trace

    C:\Temp>tshark -r P-Inside-sliced.pcapng -o tcp.calculate_timestamps:TRUE -Y tcp.dstport==443 -qz io,stat,0,SUM(tcp.time_delta)tcp.time_delta

============================================
| IO Statistics                            |
|                                          |
| Duration: 18.2 secs                      |
| Interval: 18.2 secs                      |
|                                          |
| Col 1: SUM(tcp.time_delta)tcp.time_delta |
|------------------------------------------|
|              |1           |              |
| Interval     |     SUM    |              |
|---------------------------|              |
|  0.0 <> 18.2 | 106.856631 |              |
============================================

C:\Temp>

The believe that the Total Time encompassed in this trace is 18.2s ... but then, how can the Sum of tcp.time_delta be ~106s?

And further, if I sum tcp_time.delta in the other direction (notice the change from tcp.dstport to tcp.srcport) ... I would have predicted that the SUM would be some other number ... not ~106s again

    C:\Temp>tshark -r P-Inside-sliced.pcapng -o tcp.calculate_timestamps:TRUE -Y tcp.srcport==443 -qz io,stat,0,SUM(tcp.time_delta)tcp.time_delta

============================================
| IO Statistics                            |
|                                          |
| Duration: 18.2 secs                      |
| Interval: 18.2 secs                      |
|                                          |
| Col 1: SUM(tcp.time_delta)tcp.time_delta |
|------------------------------------------|
|              |1           |              |
| Interval     |     SUM    |              |
|---------------------------|              |
|  0.0 <> 18.2 | 106.856631 |              |
============================================

C:\Temp>

See http://www.skendric.com/ask_wireshark for the pcaps

BTW: I have taken two pcaps, one from just inside a particular network device, the other from just outside same device, and I'm trying to use this technique to measure the device's impact on performance

--sk