FAST TCP
Anwis Das
Ajay Gulati
Slides adapted from : IETF presentation slides
Link: http://netlab.caltech.edu/FAST/index.html
Background
Last time we saw reliable communication
requires a transport level protocol
But user’s are selfish, leading to
congestion
Van Jacobson solved the problem by
adapting TCP
 Slow Start
 Congestion Avoidance
 Accurate Estimation of RTT, etc
Is the solution still valid considering
today’s technology??
Hypothesis
TCP Reno is a good solution for low
speed networks but not a viable solution
for high-speed networks
 Too conservative, not stable, and requires
extremely small equilibrium loss probability
There exists an equation that governs
the flow of all variants of TCP
 FAST provides a solution to this equation and
shows that it is stable and reaches
equilibrium quickly
Philosophy of Reno
 Packet level
 Designed and implemented first
 Flow level
 Understood afterwards
 Flow level dynamics determines
 Equilibrium: performance, fairness
 Stability
 Design flow level equilibrium & stability
 Implement flow level goals at packet level
Problems with TCP Reno
 Equilibrium problem
 Packet level: AI too slow, MI too drastic
 Flow level: required loss probability too
small
 Dynamic problem
 Packet level: must oscillate on binary
signal
 Flow level: unstable at large window
TCP Flow Dynamics
Reno TCP
 Packet level
W

W + 1/W
Loss: W

W – 0.5W
ACK:
 Flow level dynamics
pkts
TCP Variant Performance
 Reno: AIMD (1, 0.5)
ACK:
W  W + 1/W
Loss: W  W – 0.5W
 HSTCP: AIMD (a(w), b(w))
ACK:
W  W + a(w)/W
Loss: W  W – b(w)W
 STCP: MIMD (1/100, 1/8)
ACK:
W  W + 0.01
Loss: W  W – 0.125W
Flow level:
Reno, HSTCP, STCP, FAST
 Generic flow level dynamics!
window
adjustment
=
control
gain
flow level
goal
 Different gain k and utility Ui
 They determine equilibrium and stability
 Different congestion measure pi
 Loss probability (Reno, HSTCP, STCP)
 Queueing delay (Vegas, FAST)
Problem: binary signal
TCP
oscillation
Solution: multibit signal
FAST
stabilized
FAST TCP Dynamics
 Generic equation:
window
adjustment
=
control
gain
flow level
goal
 FAST TCP equation:
 Can map generic equation to FAST
TCP equation and understand how
FAST TCP stabilizes faster than
other TCP variants
Window control algorithm
 Full utilization
 regardless of bandwidth-delay product
 Globally stable
 exponential convergence
 Fairness
 weighted proportional fairness
 parameter a
Window control algorithm
Theorem (Jin, Wei, Low ‘03)
In absence of delay
 Mapping from w(t) to w(t+1) is contraction
 Global exponential convergence
 Full utilization after finite time
 Utility function: ai log xi (proportional fairness)
FAST
Linux
Dynamic sharing: 3 flows
Steady throughput
HSTCP
STCP
queue
FAST
loss
Linux
throughput
30min
Dynamic sharing on Dummynet
 capacity = 800Mbps
HSTCP
 delay=120ms
 14 flows
 iperf throughput
 Linux 2.4.x (HSTCP: UCL)
STCP
queue
FAST
loss
Linux
throughput
HSTCP
HSTCP
30min
STCP
Aggregate throughput
large
window
8000
small
window
800pkts
Dummynet: cap = 800Mbps; delay = 50-200ms; #flows = 1-14; 29 expts
Fairness
Jain’s index
Dummynet: cap = 800Mbps; delay = 50-200ms; #flows = 1-14; 29 expts
Stability
stable in
diverse
scenarios
Dummynet: cap = 800Mbps; delay = 50-200ms; #flows = 1-14; 29 expts
Open issues
 1 BaseRTT estimation
 route changes, dynamic sharing
 does not upset stability
 2 Small network buffer
 at least like TCP
 adapt a on slow timescale, but how?
 3 TCP-friendliness
 friendly at least at small window
 tunable, but how to tune?
 4 Reverse path congestion
 should react? rare for large transfer?
 5 Did not compare to Vegas
Future plans
What people on the street are
saying about FAST TCP…
 PC Magazine“The fastest hardware in the world will
not do you much good if your software can't take
advantage of it.
Business Week “Connections that leave
broadband in the dust”
 Steven Low “This is OK for driving slowly in a
parking lot. But on the open road you need to be
able to look further ahead: That's what we are
doing with Fast TCP.”
 Future Zone “Scientists have developed a new data
transfer protocol for the Internet fast enough to
download a full-length DVD movie in less than five
seconds”
And Of Course…
 Anwis and Ajay (Rice CS) – “FAST TCP
Rocks!!! We can’t wait to use it!”