Mitigating E Egregi egiou ous A ACK CK Delays in Ce n Cellular - - PowerPoint PPT Presentation

Mitigating E Egregi egiou

• us A

ACK CK Delays in Ce n Cellular Da Data Ne Networks b by Eliminating T TCP A CP ACK Cl CK Cloc

• cking

Wai Kay Leong, Yin Xu, Ben Leong, Zixiao Wang

National University of Singapore

Asymmetry i in C Cellula lar r Network rks

• Congestion in Uplink

‒ Concurrent Upload

▪ RSFC [Xu et al. 2012]

‒ Other Users

Downlink Uplink

Egregi giou

• us A

ACK D CK Delays

• TCP congestion control is ACK-

clocked

Data ACK Data

Egregi giou

• us A

ACK D CK Delays

• TCP congestion control is ACK-

clocked

• Congested uplink can delay ACKs

‒ Downlink becomes idle

Data ACK Data

Egregi giou

• us A

ACK D CK Delays

• TCP congestion control is ACK-

clocked

• Congested uplink can delay ACKs

‒ Downlink becomes idle

Data ACK

Idle Downlink Under Utilization Reduced Throughput

Sol

• lut

ution

• n: E

Eliminate A ACK Cl CK Cloc

• cki

king

Idea: If we know the bandwidth,

we can send at maximum rate.

Bandwidth: X Mb/s ACK Not dependent on ACK arrival Data Data Send Data at X Mb/s Data

cwnd

Cha Challenge 1 e 1: Estimating Ban Bandwidth

Bandwidth ≡ Receive Rate Idea

• Use receiving rate as equivalent of available

bandwidth

Done Passively Condition

• To avoid modifications at the receiver

Use TCP Timestamps Solution

• Enabled by default on Android and iPhones

Estimating R Rec ecei eive R e Rate

TSval = tr0

Sender Mobile Receiver

TSval = tr0 TSval = tr1 TSval = tr1 tr1 – tr0 = Δt

ΔACK/Δt=ρ

Receive Rate ρ

Chal Challenges es

• 1. Estimating Bandwidth
• 2. Timestamp Granularity too

Coarse

‒ Cannot estimate with high accuracy

• 3. Bandwidth variation

‒ Have to keep updating estimation

Sol

• lut

ution

• n

Self-oscillating Feedback Loop

‒ Estimate Receive Rate ρ ‒ Send Rate σ

Send Faster ( σ > ρ ) Send slower ( σ < ρ )

No Congestion Link Congested

How to detect congestion?

Detect ect Cong Conges estion

• n
• Idea: Monitor Queuing Delay
• How?

‒ TCP Timestamps ‒ Relative Difference between sender and receiver Buffer

tbuff > 0 tbuff = 0

Detect ecting Cong g Conges estion

Relative Delay (RD) Min Delay (RDmin)

Sender Mobile Receiver

x x

Queuing Delay (tbuff)

TSval = tr1 TSval = tr1 ts1

A

RD = tr1 – ts1

RD – RDmin ≈ tbuff

Congestion detected when tbuff >T

Sum ummary of

• f Al

Algor

• rithm
• Send 10 packets
• Estimate ρ using

replies

• 1. Initial Receive

Rate Estimation

• 2. Buffer

Management Mode

Send Faster (σ >ρ) a) Buffer Fill State Send Slower (σ <ρ) b) Buffer Drain State

• Probe network
• Details in paper
• 3. Monitor Mode

tbuff > T tbuff < T

Significant changes in network

TCP-RRE

(Receiver-Rate Estimation)

Para rameters rs?

• How much faster or slower to

send?

• What threshold T to use?
• When to switch to monitor state?

See details in paper

ns-2 E 2 Evaluation

Measured real networks to get simulation parameters

ns-2 E 2 Evaluation

1. Single Download with Slow Uplink 2. Single Download under Normal Conditions 3. Download with Concurrent Upload 4. Handling Network Fluctuation 5. TCP Friendliness

ns-2 E 2 Evaluation

1. Single Download with Slow Uplink 2. Single Download under Normal Conditions 3. Download with Concurrent Upload 4. Handling Network Fluctuation 5. TCP Friendliness

Do Download w with S Slow Up Uplink

Data ACK Downlink Speed (Mb/s) Uplink Speed (kb/s) 1 200 20 8 1MB

ns-2

Do Download w with S Slow Up Uplink

Uplink speed decreases ns-2

Do Downloa

• ad u

unde der N Normal C Condi ditions

• ns

ns-2

Do Downloa

• ad u

unde der N Normal C Condi ditions

• ns

Packet Loss

ns-2

1400+

Do Downloa

• ad u

unde der N Normal C Condi ditions

• ns

ns-2

Download w with C Con

• ncurrent U

Uploa

• ad

Downlink Speed (Mb/s) Uplink Speed (kb/s) 0.5 3000 250 8 Download 1MB Upload until download completion

ns-2

Download w with C Con

• ncurrent U

Uploa

• ad

ns-2 2.5X

Eval aluation

• n i

in n Linux

• Several Places, different ISPs
• Multiple times
• CDF of all experiments at each place

for each ISP

Upload until download completion Download 1MB with TCP-RRE Download 1MB with CUBIC

Eval aluation

• n i

in n Linux

Eval aluation

• n i

in n Linux

+50%

Eval aluation

• n i

in n Linux

Eval aluation

• n i

in n Linux

+25%

Conc Conclusion

• n

Congested Uplink Delayed ACKs Poor Download Performance

Conc Conclusion

• n
• TCP-RRE

‒ ACK Clocking ‒ Rate Control with Feedback Loop

• Use TCP Timestamp

‒ Estimate Receive Rate ‒ Detect Congestion

• Improves TCP

‒ Uplink is Slow ‒ Uplink is Congested

• Keep the Delay Low
• Fair to Other TCP Flows

Than ank Y You

• u

QUE QUESTI TIONS

Han andling g Network F k Fluct ctuation

• ns

CUBIC

Han andling g Network F k Fluct ctuation

• ns

TCP-RRE

TCP F CP Friendliness

• Run two RSFC uploads

concurrently

• Calculate Jain fairness index:

)) ( 2 /( ) (

2 2 2 1 2 2 1

R R R R + +

TCP F CP Friendliness