When I wrote the conversion script from ICMP Ping to the replayer's delay traces, I encountered a decision which is that: the slot division of replayer delay traces is 10 ms apart, but real ICMP ping traces could have RTT responses with irregular timing.
For example:
- Normal cases. RTT responses come at a 10 ms interval. We can just round the timestamps of responses and fit into consecutive rows of delay traces. -> NO PROBLEM.
- Delayed cases. RTT responses with consecutive seq nums arrive longer than the 10 ms interval. -> How do we interpolate the 10 ms time slot in between two consecutive ICMP responses?
- Loss cases. RTT responses with a seq num jump arrive longer than the 10 ms interval. -> How do we interpolate the 10 ms time slot in between two ICMP ping responses?
Also consider you have applied a loss layer for the environment which RANDOMLY drop packets with a given avg loss rate (mentioned in the README of replayer), how could you reflect the packet losses specifically due to the global reconfigurations?
That is why we ask this RTT interpolation question. Because we think to accurately reflect the losses due to the global reconfigurations, we might need to disable the loss shell, and assign high RTTs for time slots due to losses.
Look forward to your comments.
Best,
Sizhe
When I wrote the conversion script from ICMP Ping to the replayer's delay traces, I encountered a decision which is that: the slot division of replayer delay traces is 10 ms apart, but real ICMP ping traces could have RTT responses with irregular timing.
For example:
Also consider you have applied a loss layer for the environment which RANDOMLY drop packets with a given avg loss rate (mentioned in the README of replayer), how could you reflect the packet losses specifically due to the global reconfigurations?
That is why we ask this RTT interpolation question. Because we think to accurately reflect the losses due to the global reconfigurations, we might need to disable the loss shell, and assign high RTTs for time slots due to losses.
Look forward to your comments.
Best,
Sizhe