Closed yangfurong closed 2 years ago
In the original paper, they use v/(d*cwnd_in_pkts)
to update cwnd for every single acked pkt. If the cwnd is in bytes and the update of cwnd is called for multiple acked pkts, we could multiply the numerator and denominator by MSS. Then we get the adjustment component as (v * total_acked_bytes) / (d * cwnd_in_bytes)
, where total_acked_bytes
should be total_acked_pkt * MSS
.
Oh. Sorry. I found this is already a closed issue. Now I understand.
Hi. I have found the code that you use to calculate the congestion window adjustment component may be problematic.
Is it correct to multiply
conn_.udpSendPacketLen
twice?