Title: Equilibrium
1Equilibrium Stability of TCP
- Steven Low
- CS/EE Caltech
- netlab.caltech.edu
- June 2002
2Outline
- Equilibrium of TCP
- Stability of TCP
- FAST project overview
- Acknowledgments
- D. Choe (Postech/Caltech), J. Doyle (Caltech), C.
Jin (Caltech), K. Kim (SNU/Caltech), F. Paganini
(UCLA), J. Wang (Caltech) - L. Peterson, L. Wang (Princeton)
- S. Athuraliya, D. Lapsley (UMelb)
3Congestion Control
- Heavy tail ? Mice-elephants
4TCP AQM
pl(t)
xi(t)
- Example congestion measure pl(t)
- Loss (Reno)
- Queueing delay (Vegas)
5TCP/AQM
- Congestion control is a distributed asynchronous
algorithm to share bandwidth among users - It has two components
- TCP adapts sending rate (window) to congestion
- AQM adjusts feeds back congestion information
- They form a distributed feedback control system
- Equilibrium stability depends on both TCP and
AQM - And on delay, capacity, routing, connections
6Understanding TCP/AQM
- Protocol (Reno, Vegas, RED, REM/PI)
7Equilibrium duality model
- TCP/AQM
- Maximize aggregate source utility
- With different utility functions
8Equilibrium of Vegas
- Network
- Link queueing delays pl
- Queue length clpl
- Sources
- Throughput xi
- E2E queueing delay qi
- Packets buffered
- Utility funtion Ui(x) ai di log x
- Proportional fairness
9Stability of AIMD/RED
- Small effect on queue
- AIMD
- Mice traffic
- Heterogeneity
- Big effect on queue
- Stability!
10Stable 20ms delay
Window
Ns-2 simulations, 50 identical FTP sources,
single link 9 pkts/ms, RED marking
11Stable 20ms delay
Window
Ns-2 simulations, 50 identical FTP sources,
single link 9 pkts/ms, RED marking
12Unstable 200ms delay
Window
Ns-2 simulations, 50 identical FTP sources,
single link 9 pkts/ms, RED marking
13Unstable 200ms delay
Window
Ns-2 simulations, 50 identical FTP sources,
single link 9 pkts/ms, RED marking
14Other effects on queue
20ms
200ms
15Stability region
- Unstable for
- Large delay
- Large capacity
- Small load
16Current protocols
- Current TCP/RED has equilibrium and stability
problems - Equilibrium problems
- Unfairness to connections with large delay
- At high bandwidth, equilibrium loss probability
too small to be reliable - Stability problems
- Unstable as delay, or more strikingly, as
capacity scales up - Instability causes large slow-timescale
oscillations - Long time to ramp up after packet losses
- Jitters in rate and delay
- Underutilization as queue jumps between empty
high
17FAST Fast AQM Scalable TCP
- FAST solves these equilibrium and stability
problems - Equilibrium properties
- Uses end-to-end delay rather than loss as
congestion measure - Achieves any desired fairness, expressed by
utility function - Very high utilization (99 in theory)
- Stability properties
- Maintains stability for arbitrary delay,
capacity, routing load - FAST rescales itself as these parameters change
- Non-oscillatory in equilibrium
- Need not re-tune as these parameters change
- Good performance
- Negligible queueing delay loss
- Fast response
18Network upgrade 2001-06
19Projected performance
04 5
05 10
Ns-2 capacity 155Mbps, 622Mbps, 2.5Gbps,
5Gbps, 10Gbps 100 sources, 100 ms round trip
propagation delay
J. Wang (Caltech)
20Projected performance
TCP/RED
FAST
Ns-2 capacity 10Gbps 100 sources, 100 ms round
trip propagation delay
J. Wang (Caltech)
21Coming together
Clear present Need
Resources
22Coming together
Clear present Need
FAST Control
Resources
23Papers
netlab.caltech.edu
- Optimization flow control, I basic algorithm
convergence (ToN, 7(6), Dec 1999) - A duality model of TCP flow controls (ITC, Sept
2000) - Understanding Vegas a duality model (J. ACM,
2002) - Scalable laws for stable network congestion
control (CDC, 2001) - Stabilized Vegas (Allerton Conf, 2002)