RTT Measurement of TCP
Impact of RTT-gain coefficient
- Acts as „smoothing“ factor
- Considered values
Results
- Derivation less than 1ms
- „Smoothed“ curved closerto real round-trip delayvariation
Notes:
Simulator Settings:
timer granularity: none
min. RTT valu: 1ms
==> smoothed timer gives better approximation of RTT variation
==> still suboptimal estimation of „real“ RTT due to TCP contrains
transmission window experinces havy changes
<=> needless retransmissions due to low timer granularity
queuing delay adds a constant offset of approx. 38ms
Inlcude later:
picture of queuing delay & transmission window size
can we reduce the queuing delay with very hight speed link
==> should only work if BDP is larger than window size