Are simgrid predictions pessimistics for small packets ?
throwing some notes about some observed predictions on the inflight duration of packets
with 20ms latency.
coordinator logs
# End of previous turn.
- No message has been through simgrid yet (next_event=inf)
[nova-0.lyon.grid5000.fr:vm_coordinator:(1) 42.020000] /home/msimonin/workspace/repos/2018-vsg/src/simgrid/VmsCoordinator.cpp:141: [vm_coordinator/DEBUG] next deadline = 42.040000 [time+min_latency=42.040000, next_reception_time=inf]
[nova-0.lyon.grid5000.fr:vm_coordinator:(1) 42.020000] /home/msimonin/workspace/repos/2018-vsg/src/simgrid/VmsInterface.cpp:150: [vm_interface/DEBUG] Sending: go to deadline 42.040000 (42.040000)
[nova-0.lyon.grid5000.fr:vm_coordinator:(1) 42.020000] /home/msimonin/workspace/repos/2018-vsg/src/simgrid/VmsInterface.cpp:162: [vm_interface/DEBUG] getting the message send by the VMs
# New turn
[nova-0.lyon.grid5000.fr:vm_coordinator:(1) 42.020000] /home/msimonin/workspace/repos/2018-vsg/src/simgrid/VmsInterface.cpp:150: [vm_interface/DEBUG] Sending: go to deadline 42.040000 (42.040000)
## An icmp has been captured
[nova-0.lyon.grid5000.fr:vm_coordinator:(1) 42.020000] /home/msimonin/workspace/repos/2018-vsg/src/simgrid/VmsInterface.cpp:182: [vm_interface/VERBOSE] getting a message from VM 192.168.122.11
[nova-0.lyon.grid5000.fr:vm_coordinator:(1) 42.020000] [vm_interface/INFO] got the message [RT] (size 98) from VM [192.168.122.11](=ip(192.168.122.11)) to VM [175810752](=ip(192.168.122.10)) with timestamp [42.022235]
[nova-0.lyon.grid5000.fr:vm_coordinator:(1) 42.020000] /home/msimonin/workspace/repos/2018-vsg/src/simgrid/VmsInterface.cpp:218: [vm_interface/DEBUG] forwarding all the 1 messages to SimGrid
## Inject the message into the network model
[nova-0.lyon.grid5000.fr:vm_coordinator:(1) 42.020000] /home/msimonin/workspace/repos/2018-vsg/src/simgrid
/VmsCoordinator.cpp:153: [vm_coordinator/DEBUG] going to time 42.022235
[nova-2.lyon.grid5000.fr:sender:(4) 42.022235] [vm_coordinator/INFO] sending (size 98) from vm [192.168.122.11], to vm [192.168.122.10] (on pm [nova-1.lyon.grid5000.fr])
[nova-0.lyon.grid5000.fr:vm_coordinator:(1) 42.040000] /home/msimonin/workspace/repos/2018-vsg/src/simgrid
## Preparing next turn
- the pending com is taken into account in the deadline calculation
- the message will be delivered at 42.282435 which is 260ms later (on a 20ms latency link).
/VmsCoordinator.cpp:194: [vm_coordinator/DEBUG] Timestep finished preparing the next iteration [current_time=42.040000] [next_event = 42.282435]
Qemu console output
note: 2 x 260 = 520
Notes
The same delay can be observed in the send example (without qemus)