!!!!!!!

TCP$Alterna,ve$Backoff$with$ECN$(ABE)$

Naeem!Khademi,!Michael!Welzl,!Grenville!Armitage,!Chamil!Kulatunga,!David!Ros,!Gorry!Fairhurst,!Stein!Gjessing!and!SebasBan!Zander$

!!

TCPM!WG!F!IETF!93,!Prague!–!July!22,!2015!naeemk@ifi.uio.no!

!

•  With$standard$TCP$CC’s$mul,plica,ve$decrease$factor$(β!=!0.5):$more!throughput!means!more!buffering!(more!latency)!

–  1!BDP!for!full!uBlizaBon!!!

•  AQM$solu,ons:$aiming!to!keep!the!average!latency!low:!(FQ_)CoDel,!PIE,!ARED,!…!!!

•  AQMs’!default!thresholds!translate!into!a!Bny!average!buffer!(e.g.!5ms~20ms!for!CoDel!and!PIE)!

!

Internet’s$Latency$Problem$

2$

7.6 7.8

8 8.2 8.4 8.6 8.8

9 9.2 9.4 9.6 9.8 10

10 20 30 40 50 60 70 80 90 100

Thro

ughp

ut (M

bps)

Buffer Size (pkts)

BDP

2/3 BDP

3/7 BDP

1/4 BDP

1/9 BDP

β=0.5β=0.6β=0.7β=0.8β=0.9

SimulaBon!vs.!model!One!NewReno!flow!@!10!Mbps,!RTT!100ms,!DropTail!!!

!!!!

!!•  UnderFuBlizaBon!sBll!a!problem!when!RTT!increases!

The$Problem$with$using$AQMs$

3$

●

●

●●

●

●

●

●

●

●

●

●

●

●

●

●

●●●●

●●●●

●

●

●

●●

●

●

●

●

●●●

●●

●●

●

●

●●

●

●

●

●

●

Thro

ughp

ut (M

bps) ●

●

●●

●

●

●

●

●

●

●

●

●

●

●

●

●●●●

●●●●

●

●

●

●●

●

●

●

●

●●●

●●

●●

●

●

●●

●

●

●

●

●

20 40 80 160 240RTT (ms)

10

12

14

16

18

20

No_ECN ECN

CoDel

●●

●

●●●●●●●●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●

●

●

●●

●

●

●

●

●

●

●●●●●●●●●●

●

●●

●

●

●

●

●

●

●

●

●

●

●

●

●●●●●●●●●

●

●

●

●

●

●

●

●

●

●●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●●●●●●●

●

●

●

●

●●

●

●

●

●

●

●

●●●●

●

●

●

●●

●

●

●

●

●

●

●●

●

●

●

●

●

●

●●

●

●●

●

●

●●

●

●

●

●

●

●●

●●

●●

●

●

●

●

●

●

●

●●●

●●

●●

●●

Thro

ughp

ut (M

bps)

●●

●

●●●●●●●●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●

●

●

●●

●

●

●

●

●

●

●●●●●●●●●●

●

●●

●

●

●

●

●

●

●

●

●

●

●

●

●●●●●●●●●

●

●

●

●

●

●

●

●

●

●●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●●●●●●●

●

●

●

●

●●

●

●

●

●

●

●

●●●●

●

●

●

●●

●

●

●

●

●

●

●●

●

●

●

●

●

●

●●

●

●●

●

●

●●

●

●

●

●

●

●●

●●

●●

●

●

●

●

●

●

●

●●●

●●

●●

●●

20 40 80 160 240RTT (ms)

10

12

14

16

18

20

No_ECN ECN

PIE

One CUBIC flow @ 20 Mbps (real-life tests)

!!  Link!underFuBlizaBon!with!AQMs’!

default!thresholds!when!RTTs!head!above!60ms~80ms!!!

!

Problem$with$longLRTT$paths$

link capacity(C.2Tp)

time

Under-utilized portion of the link capacity

buffering (b)

CWmax

β.CWmax

!

!

Can$we$use$a$larger$TCP$β?$$

4$

150

200

250

300

350

0 10 20 30 40 50 60

cwnd

(pa

cket

s)

Time (sec)

Capacity limit

RTTmin 213ms

RTTmin 283ms

Queue length: 1/2 BDPQueue length: 1 BDP

•  Packet$loss:$tailFloss!(and!other!sources!of!loss)!versus!AQMFloss!!–  Not!a!good!indicator!of!the!onset!of!congesBon!

•  An!ECNLCE$signal!explicitly!indicates!an!AQM’s!presence!(most!likely!to!be!(FQ_)CoDel!or!PIE)!with!a!low!marking!threshold!!!

•  Alterna,ve$Backoff$with$ECN$(ABE):$only$a!simple!TCP!parameter!change!=>!use!a!β!higher!than!0.5!in!response!to!ECNFCE!(βECN)!and!react!with!β=0.5!in!response!to!loss!!!

–  Few!lines!of!code!in!kernel!(in!the!sender)only)!

!

Alterna,ve$Backoff$with$ECN$(ABE)$

5$

Latency$vs.$throughput$tradeoff$(#1)$

6$

0 50 100 150 200 250 300

010

2030

4050

6070

80

RTT=100ms

Byte

s D

epar

ted

βecn

0.60.70.750.8

0.850.90.95

Impr

ovem

ent o

ver β

ecn=

0.5

(%)

0 50 100 150 200 250 300

RTT=200ms

Time(s)0 50 100 150 200 250 300

RTT=400ms

0 2 4 6 8 10

020

4060

8010

0 RTT=100ms

CD

F (%

)

βecn

0.50.60.70.750.80.850.90.95

CD

F(%

)

0 2 4 6 8 10

RTT=200ms

Queueing Delay (ms)0 2 4 6 8 10

RTT=400ms

One NewReno flow, CoDel @ 10 Mbps, RTT=100

Latency$vs.$throughput$tradeoff$(#2)$

7$

One NewReno flow, PIE @ 10 Mbps, RTT=100

0 50 100 150 200 250 300

020

5080

110

140

170

200

RTT=100ms

Byte

s D

epar

ted

βecn

0.60.70.750.8

0.850.90.95

Impr

ovem

ent o

ver β

ecn=

0.5

(%)

0 50 100 150 200 250 300

RTT=200ms

Time(s)0 50 100 150 200 250 300

RTT=400ms

0 5 10 15 20 25 30 35

020

4060

8010

0 RTT=100ms

CD

F (%

)

βecn

0.50.60.70.750.80.850.90.95

CD

F(%

)

0 5 10 15 20 25 30 35

RTT=200ms

Queueing Delay (ms)0 5 10 15 20 25 30 35

RTT=400ms

Latency$vs.$throughput$tradeoff$(#3)$

8$8$

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

Thro

ughp

ut (M

bps)

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

400.5

800.5

1600.5

2400.5

400.6

800.6

1600.6

2400.6

400.7

800.7

1600.7

2400.7

400.8

800.8

1600.8

2400.8

RTT (ms)βecn

5

10

15

20

RTT

(ms)

400.5

800.5

1600.5

2400.5

400.6

800.6

1600.6

2400.6

400.7

800.7

1600.7

2400.7

400.8

800.8

1600.8

2400.8

RTT (ms)βecn

0

50

100

150

200

250 CoDel!RTT! Throughput!

Increased βECN causes insignificant change in median RTT

Source:!hkp://frenchyme.blogspot.no!

!!  Low$latency$and$high$u,liza,on$$

ABE$Performance$$

TCP$Convergence$Time$

9$9$

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0 1 2 3 4 5

CD

F

Convergence time / default (βecn 0.5)

βecn 0.6βecn 0.7

βecn 0.75βecn 0.8

βecn 0.85βecn 0.9

Two NewReno flows, CoDel

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0 1 2 3 4 5

CD

F

Convergence time / default (βecn 0.5)

βecn 0.6βecn 0.7

βecn 0.75βecn 0.8

βecn 0.85βecn 0.9

Two NewReno flows, PIE

@ {1, 5, 10, 20, 40, 100} Mbps and RTT={10, 20, 40, 60, 80, 160, 240, 320} ms

JFI of the cumulative number of bytes transferred per flow from the time a new flow entered, (after 30 sec during a 90 sec experiment) and noting the time it took until this index reached a threshold (0.95)

Friendliness$with$standard$$(lossLbased)$TCP$

10$10$

0.5

1

1.5

2

2.5

3

3.5

1 vs 9 3 vs 7 5 vs 5 7 vs 3 9 vs 1

Nor

mal

ized

thro

ughp

ut ra

tio(E

CN

flow

s / n

oEC

N fl

ows)

# noECN flows vs # ECN flows

βecn=0.5βecn=0.6

βecn=0.7βecn=0.75

βecn=0.8βecn=0.85

βecn=0.9

CoDel

0.5

1

1.5

2

2.5

3

3.5

1 vs 9 3 vs 7 5 vs 5 7 vs 3 9 vs 1

Nor

mal

ized

thro

ughp

ut ra

tio(E

CN

flow

s / n

oEC

N fl

ows)

# noECN flows vs # ECN flows

βecn=0.5βecn=0.6

βecn=0.7βecn=0.75

βecn=0.8βecn=0.85

βecn=0.9

PIE

Normalized throughput ratio between two groups of 10 flows (CUBIC with different βECN against standard TCP) @10 Mbps using RTT={20, 80, 160, 320} ms.

ABE$and$slowLstart$(#1)$

11$11$0 2 4 6 8 10 12

010

020

030

040

0

Time (s)

cwnd

(pac

kets

)

RTT, βecn

160ms,0.5240ms,0.5160ms,0.9240ms,0.9

160ms X 20Mbps capacity

240ms X 20Mbps capacity

The!effect!of!overshoot!at!the!end!of!slowFstart!!@!20Mbps!(experiment)!!

-  Benefits!short!flows!terminaBng!right!aner!SS!-  E.g.!reducBon!in!PLT!of!large!webFpages!-  Number!of!reducBon!steps!bound!to!logβECN (0.5)

ABE$and$slowLstart$(#2)$

12$12$

Percentage reduction in FCT (compared to using βECN = 0.5) with different flow sizes @10 Mbps link with RTT =100ms

0

5

10

15

20

25

600 800 1000 1200 1400 1600 1800 2000

Red

uctio

n in

Flo

w C

ompl

etio

n Ti

me

(%)

File Size (KB)

CoDel

0

5

10

15

20

25

600 800 1000 1200 1400 1600 1800 2000

File Size (KB)

CoDel PIEβecn=0.6βecn=0.7βecn=0.75βecn=0.8βecn=0.85βecn=0.9βecn=0.95

•  Alterna,ve$Backoff$with$ECN$(ABE):$$

Naeem!Khademi,!Michael!Welzl,!Grenville!Armitage,!Chamil!Kulatunga,!David!Ros,!Gorry!Fairhurst,!Stein!Gjessing,!SebasBan!Zander,!"Alterna,ve$Backoff:$Achieving$Low$Latency$and$High$Throughput$with$ECN$and$AQM",!Technical!report!150710A,!10!July!2015,!Swinburne!University!of!Technology,!Australia!!!hkp://caia.swin.edu.au/reports/150710A/CAIAFTRF150710A.pdf!!

dranFkhademiFalternaBvebackoffFecn!!

–  A!minor!senderFside!modificaBon!(changes!βECN)!–  Complies!with!RFC!3168!–  Incremental!deployment!with!no!flagFday!!

$

Conclusions$

13$

!!  Significant!throughput!gain!with!lightlyFmulBplexed!traffic!!  Low!latency!(using!CoDel!or!PIE)!!  No!starvaBon!of!standard!TCP!!  Reasonable!convergence!and!fairness!with!recommended!βs!

ABE$Performance$$

Q&A$

14!