Back to index
Simulation-based Comparisons of Tahoe, Reno and SACK TCP
Kevin Fall and Sally Floyd
- Main point: Nothing compares to SACK TCP!
- Tahoe TCP: Follows a basic go-back-n model using
slow-start, congestion avoidance and fast retransmit algorithms. With
Fast Retransmit, after receiving a small number of duplicate acks for
the same TCP segment, the data sender infers that the packet has been
lost and retransmits the packet without waiting for the retransmission
timer to expire.
- Reno TCP:
- Modification to the Tahoe TCP Fast Retransmit
algorithm to include Fast Recovery; this prevents the pipe from going
empty after Fast Retransmit, thereby avoiding the need to slow start
after a single packet loss.
- A TCP sender enters Fast Recovery after receiving athreshold number of
dup acks. The sender retransmits one packet and reduces its congestion
window by half. Instead of slow-starting, the Reno sender uses
additional incoming dup acks to clock subsequent outgoing packets.
- Reno TCP greatly improves performance in the face of single
packet loss, but can suffer when multiple packets are lost.
- New Reno TCP: In Reno, a partial ack (ack for some but not
all of the packets that were outstanding at the start of the fast
recovery period) takes TCP out of Fast Recovery. In New Reno, partial
acks fo not take TCP out of fast recovery; partial acks receivbed
during fast recovery are treated as an indication that the packet
immediately following the acked packet has been lost and should be
retransmitted. Thus, when multiple packets are lost, New Reno can
recover without a retransmission timeout.
- SACK TCP: The TCP sender maintain a scoreboard which
keeps track of acks from previous SACK packets. When the sender is
allowed to send a packet, it retransmits the next packet from the list
of packets inferred to be missing at the receiver. If there are no
such packets and the receiver's advertized window is sufficiently
large, the sender sends a new packet.
- One packet loss: Tahoe TCP does badly due to slow-start after
the packet loss. All others do relatively the same.
- Two packet losses: Reno TCP fails to do as well as New Reno or
SACK TCP, since its algorithm is tuned for single packet losses.
- Multiple packet losses: Reno TCP performs miserably in the face
of a large number of packet losses. SACK TCP continues to
out-perform the rest of the algorithms.
Back to index (Created by Yatin Chawathe)