I'm slightly confused here with traceroute.

I understand that times listed on row each are the round trip time from the my machine to the target host (traceroute sends 3 separate packets). I'm trying to figure out how I can get the intermediate latencies between hops (e.g. from 1 to 2 then 2 to 3 then 3 to 4) and since I need to map this into a diagram.

I thought it would be sufficient to take a row (apart from 1st), work out the average RTT on that row and then subtract that average RTT from the previous row's avg RTT to get the intermediate latency. But what if the times decrease? Surely I get negatives which can't be correct?

traceroute to www.draw.io (, 30 hops max, 60 byte packets
 1  gateway (  0.308 ms  0.270 ms  0.282 ms
 2 (  0.321 ms  0.302 ms  0.342 ms
 3 (  0.142 ms  0.130 ms  0.162 ms
 4 (  0.407 ms  0.420 ms  0.411 ms
 5  xe-5-0-0.londpg-ban1.ja.net (  0.904 ms  0.894 ms  0.863 ms
 6  ae26.londpg-sbr2.ja.net (  0.852 ms  0.926 ms  0.906 ms
 7  ae30.londtw-sbr2.ja.net (  1.423 ms  1.452 ms  1.442 ms
 8  ae29.londtn-sbr1.ja.net (  1.385 ms  1.482 ms  1.469 ms
 9  ae0.londtn-ban3.ja.net (  1.662 ms  1.691 ms  1.682 ms
10  linx-juniper.as13335.net (  1.537 ms  1.508 ms  1.505 ms
11 (  1.553 ms  1.527 ms  1.547 ms

Would hugely appreciate some help here.


When finely measured, Network latency varies widely millisecond-by-millisecond, and each measurement is taken seconds apart, so you cannot get the answer you want as the result of a single measurement of each hop.

You will probably be best off repeating the test a number of times, and recording the times for each hop. Then you can do your analysis based on the mean (average) for a given hop, or get more sophisticated with it, calculating ranges for standard deviations, etc. that will also let you throw out outlier values that will dramatically affect the results.

  • Understood but I'm bound to still get cases where an RTT is smaller than the row above. How can I handle that? I need to somehow get the intermediate latencies. – ak1652 Nov 1 '16 at 13:17
  • you should be able to throw out the outlying values where the RTT is less than the prior hops RTT. You should find that there are fewer cases where that occurs than cases where it does. If not, then either your goal-state is unreachable (eg there is physically no way to tell based on your conception of incremental latency) or tracert is simply an inappropriate tool for the job. It makes sense that if more of your readings are impossible than possible, that you are either asking the wrong question, or that you lack to tools to obtain an answer to it. – Frank Thomas Nov 1 '16 at 13:44

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.