Cubic: Qian HE (Steve) CS 577 - Prof. Bob Kinicki

CUBIC
Qian HE (Steve)
CS 577 – Prof. Bob Kinicki

Agenda
• Brief Introduction of CUBIC
• Prehistory of CUBIC
– Standard TCP
– BIC
• CUBIC
• Conclusion
2
Brief Introduction
• CUBIC is a less aggressive and more

systematic derivative of BIC, in which the
window is a cubic function of time since the
last congestion event, with the inflection
point set to the window prior to the event.
3
Why do we need CUBIC-TCP?
• Compares to:
– Standard TCP
– BIC-TCP
4
Standard TCP
• Underutilization of the bandwidth in High-Speed Network
L ink U tilization
1 Cannot fully utilize the

0.9 huge capacity of high-
0.8 speed networks!
0.7
0.6
0.5
0.4
0.3
0 2000 4000 6000 8000 10000
Link Capacity (Mbps)
NS-2 Simulation (100 sec)

 Link Capacity = 155Mbps, 622Mbps, 2.5Gbps, 5Gbps, 10Gbps,
 Drop-Tail Routers, 0.1BDP Buffer
 5 TCP Connections, 100ms RTT, 1000-Byte Packet Size
5
Presentation: "Congestion Control on High-Speed Networks”, Injong Rhee, Lisong Xu, Slide 6
Standard TCP
• Low window size resilience to packet loss in High-Speed Network
1.4 hours 1.4 hours 1.4 hours
Packet loss Packet loss Packet loss Packet loss TCP

cwnd 100,000 10Gbps
50,000 Slow Increase Fast Decrease

5Gbps
cwnd = cwnd + 1 cwnd = cwnd * 0.5
Slow start Congestion avoidance Time (RTT)
Presentation: "Congestion Control on High-Speed Networks”, Injong Rhee, Lisong Xu, Slide 7 6
Why BIC?
• Existing schemes have a severe RTT
unfairness problem
• RTT unfairness for high-speed networks
occurs distinctly with drop tail routers for
flows with large congestion windows where
packet loss can be highly synchronized.
7
BIC
• “Binary Increase Congestion Control (BIC) for

Fast Long-Distance Networks”, Lisong Xu,
Khaled Harfoush, and Injong Rhee, IEEE
INFOCOM 2004
8
Goals of BIC
• Scalability: BIC can scale its bandwidth share to 10 Gbps around 3.5e-8 loss
rates (comparable to HSTCP which reaches 10Gbps at 1e-7).
• RTT fairness: for large windows, BIC’s RTT unfairness is proportional to the
inverse square of the RTT ratio as in AIMD.
• TCP friendliness: BIC achieves bounded TCP fairness for all window sizes.
Around high loss rates where TCP performs well, its TCP friendliness is
comparable to STCP’s.
• Fairness and convergence: compared to HSTCP and STCP, BIC achieves better
bandwidth fairness over both short and long time scales, and faster convergence
to a fair bandwidth share.
9
BIC Algorithm
• If cwnd < low_window, • Else, BIC

normal TCP:
– ACK received
• cwnd = cwnd + 1
– Enter recovery
• cwnd = cwnd * 0.5
10
BIC Algorithm
• Recovery
• ACK received
– If cwnd < Wmax
– If cwnd < Wmax
• Wmax = cwnd * (1 – ß / 2)
• cwnd += (Wmax – cwnd) / 2
– Else
– Else
• Wmax = cwnd
• cwnd += cwnd - Wmax
– cwnd *= 1 - ß
Smin <= cwnd
&&
cwnd <=Smax 11
BIC with no lost
Smin
256
224
192
cwnd
160
128 Linear Search
96 Binary Search with Smax and
64 Smin
32 Smax
0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Time (RTT)
12
"Binary Increase Congestion Control (BIC) for Fast Long-Distance Networks", Lisong Xu, Khaled
Harfoush, and Injong Rhee 13
Why CUBIC?
• Window control of BIC is so complex!
• BIC’s growth function can still be too

aggressive for TCP, especially under short RTT
or low speed networks.
• BIC still has room for improving TCP-
friendliness and RTT-fairness!
14
CUBIC Algorithm
• ACK received • Recovery
– Update K with:
3
cwnd =C ×t-
( K ) +Wmax
– C is a scaling factor
K = 3 b ×Wmax / C
– t is the elapsed time from the last – Update Wmax with:
Wmax =b ×Wmax
cwnd cannot
window be less than
reduction
1- b t
cwnd =b ×Wmax + 3× ×
– Wmax is the window1+ b RTT
size just before
as
thetolast
keep the growth
window rate
reduction
the same as standard TCP
– Kinisshort
updated
RTTatnetworks.
the time of last lost – β is a constant multiplication decrease
event factor
15
CUBIC window curves with competing flows (NS simulation in a network with 500Mbps and 100ms RTT), C = 0.4, β = 0.8. 16
Window Growth Function
CUBIC
BIC
"CUBIC: A New TCP-Friendly High-Speed TCP Variant", Injong Rhee, and Lisong Xu 17
Stability
4 flows of a high-speed TCP variant over a long-RTT network path (~220ms)

4 flows of long-term TCP-SACK flows over a short-RTT path (~20ms)
Coefficient of Variation (CoV)
• “There is no well-defined metric of stability.”
s
CoV
• “Often the CoV of transmission rates are
=
used to depict stability.”
m
• “For a less satisfactory measure, we plotted
the CoV of throughput.”
CoV - 20% BDP
CoV - 200% BDP
Thanks

Cubic: Qian HE (Steve) CS 577 - Prof. Bob Kinicki

Uploaded by

Copyright:

Available Formats

Cubic: Qian HE (Steve) CS 577 - Prof. Bob Kinicki

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Cubic: Qian HE (Steve) CS 577 - Prof. Bob Kinicki

Uploaded by

Copyright:

Available Formats

CUBIC

CS 577 – Prof. Bob Kinicki

• CUBIC is a less aggressive and more

1 Cannot fully utilize the

NS-2 Simulation (100 sec)

1.4 hours 1.4 hours 1.4 hours

Packet loss Packet loss Packet loss Packet loss TCP

50,000 Slow Increase Fast Decrease

Slow start Congestion avoidance Time (RTT)

• “Binary Increase Congestion Control (BIC) for

• If cwnd < low_window, • Else, BIC

• BIC’s growth function can still be too

4 flows of a high-speed TCP variant over a long-RTT network path (~220ms)

You might also like