2

Is there a (optimally free) tool to measure One-Way-Delay?

The use case is checking whether network traffic delayed by f.ex. tc is flowing as it should.

At the minimum, the application should be called with the IP address of one host, and report back the delays in each direction. It is totally acceptable to start a server on the destination host.

It should run on (at least virtualized) Linux. It would help if it were free.

There is f.ex. a RFC from 2016 dealing with this as well as this cs.SX question. Any tools that implement it or other delay-measuring tools?

Note: ping and traceroute are no valid answers, they do not measure one-way-delay.

0

The problem is that there are two different clocks running at each end the time stamp at the other end will always be to some degree out of step with the time stamp at the sending end. As you say tracert effectively does a round trip to each node that the message takes along the route as each node replies to the originator as well as passing on the message.

Unless you can connect both ends directly to a time standard such as the atomic clock that has a guaranteed synchronised time the best that you are likely to be able to do is to exchange time values between the two machines multiple times and establish the variability in the differences, i.e. if on one occasion the difference between the two times is 1.03 milliseconds and the next it is 3.02 milliseconds then either one machine has a poor clock &/or message handling queue especially likely if one or both are Windows machines or the network delay has changed.

Not really an answer but too long for a comment.

  • How about for two Linux hosts with ntp time? – M K Sep 7 '17 at 12:08
  • @MK I am reasonably sure that you will not get the required accuracy of synchronisation between clocks unless both servers are connected to the same timeserver with approximately equal lengths of cable. – Steve Barnes Sep 8 '17 at 6:54
  • How about if both are VMs running inside the same host? (question has been updated) – M K Sep 10 '17 at 10:41
  • @MK even between two VMs in a single host there is the possibility of some variation between the time to service requests for the time which would probably be of a similar order of magnitude to the delay through the virtual network between the two VMs. I also suspect that the time resolution on many platforms would be insufficient to calibrate such times. – Steve Barnes Sep 10 '17 at 14:43
  • Ok, so this answer seems to be: It won't work unless you use a special hardware setup. – M K Sep 11 '17 at 7:19

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.