When TCP Tahoe or Reno observe a loss event, the threshold gets reset to half of the last congestion window (CW). When the previous CW was an odd number, is the new threshold the floor or ceiling of CW/2
?
Asked
Active
Viewed 1,248 times
1

Joshua Swain
- 571
- 2
- 4
- 22
1 Answers
0
Both congestion window and the threshold are measured in bytes, as said in RFC 2001 :
Congestion avoidance dictates that cwnd be incremented by segsize*segsize/cwnd each time an ACK is received, where segsize is the segment size and cwnd is maintained in bytes.
and RFC 2581 :
ssthresh = max (FlightSize / 2, 2*SMSS) (3)
There is much documentation showing CWND as an integer value, it counts how many MSS, in bytes. If MSS is not an even number, I don't know as I didn't try and we typically have the same value everywhere (536 or 1460), one byte more or less this would have very little impact.

Community
- 1
- 1

Eugène Adell
- 3,089
- 2
- 18
- 34