Congestion Control in Recursive NetworkArchitectures

David Hayes Peyman Teymoori and Michael WelzlUniversity of Oslo

[davihaypeymantmichawe]ifiuionoIETF95 Buenos Aires mdash ICCRG

David IETF95 Buenos Aires mdash ICCRG 1 9

Pristine Who we are and what wersquore doing

ict-pristineeu

RINARecursive Inter-Network

Architecture

tools

RINAsimulator

Omnet++

RINA sdk

C++

Research Security

CongestionControl

Routing

Schedulingand QoS

Standards

PouzinSociety

ISO

David IETF95 Buenos Aires mdash ICCRG 2 9

The Internet and Recursive Architectures

Internet

Application

TCPUDP

IP

LinkLLCMAC

Physical

QoS MPLS etc

Proxies middle boxes etc

Recursive Architecture

Application

Generic Layer

Generic Layer

Physical

Examples includeRecursive Inter-Network Architecture (RINA)httpwwwpouzinsocietyorg

Recursive Network Architecture (RNA)httpwwwisiedurna

David IETF95 Buenos Aires mdash ICCRG 3 9

Comparative RINA Example

S1 R1 Router1

TC

P St

ack

Split

-TC

P S

tack

L1 L2 L3 L4

RIN

A

Stac

k 2-DIF

1-DIF 1-DIF

EFCP RMT

EFCP RMT

EFCP RMT

EFCP RMT

EFCP RMT

EFCP RMT

EFCP RMT

L1 L2 L3 L4

L1 L2 L3 L4

L1 L2 L3

L1 L2 L3 L4

L1 L2 L3 L4

TCP

S1 R1 Router1

TC

P St

ack

Split

-TC

P S

tack

L1 L2 L3 L4

RIN

A

Stac

k 2-DIF

1-DIF 1-DIF

EFCP RMT

EFCP RMT

EFCP RMT

EFCP RMT

EFCP RMT

EFCP RMT

EFCP RMT

L1 L2 L3 L4

L1 L2 L3 L4

L1 L2 L3

L1 L2 L3 L4

L1 L2 L3 L4

Split TCP

S1 R1 Router1

TC

P St

ack

Split

-TC

P S

tack

L1 L2 L3 L4

RIN

A

Stac

k 2-DIF

1-DIF 1-DIF

EFCP RMT

EFCP RMT

EFCP RMT

EFCP RMT

EFCP RMT

EFCP RMT

EFCP RMT

L1 L2 L3 L4

L1 L2 L3 L4

L1 L2 L3

L1 L2 L3 L4

L1 L2 L3 L4

RINA

S1 R1 Router1

TC

P St

ack

Split

-TC

P S

tack

L1 L2 L3 L4

RIN

A

Stac

k 2-DIF

1-DIF 1-DIF

EFCP RMT

EFCP RMT

EFCP RMT

EFCP RMT

EFCP RMT

EFCP RMT

EFCP RMT

L1 L2 L3 L4

L1 L2 L3 L4

L1 L2 L3

L1 L2 L3 L4

L1 L2 L3 L4

S1 R1 Router1

TC

P St

ack

Split

-TC

P S

tack

L1 L2 L3 L4

RIN

A

Stac

k 2-DIF

1-DIF 1-DIF

EFCP RMT

EFCP RMT

EFCP RMT

EFCP RMT

EFCP RMT

EFCP RMT

EFCP RMT

L1 L2 L3 L4

L1 L2 L3 L4

L1 L2 L3

L1 L2 L3 L4

L1 L2 L3 L4

From P Teymoori M Welzl G Stein E Grasa R Riggio K Rausch and D Siracuss Congestion control in the recursiveinternetwork architecture (RINA)In IEEE International Conference on Communications (ICC) Next Generation Networking and Internet Symposium May 2016

to appear

David IETF95 Buenos Aires mdash ICCRG 4 9

Congestion Control

Types of ldquosolvablerdquo Network Congestion

Network adaption ndash traffic rerouted along less congested paths

Network restriction ndash admission control and policing

End system adaption ndash traffic sources adjust send rates

Recursive CC work in RINA

Stability of chained and stacked congestion controllers

Congestion signals and how they influence performance

Stable scalable new types of congestion control

David IETF95 Buenos Aires mdash ICCRG 5 9

Recursive Congestion SignalsISP 1

AccessISP 2

Access

GWGW

Public Internet Layer

Inter ISP Layer

ISP 1 Layer ISP 2 Layer

ServerHost

Feedback options in RINA

ImplicitI Push back and push up

ExplicitI Packet marking and

reflectionI Control packet notifications

Potentially provides

Faster responses

Smaller buffer requirements

Cautions mdash Current research

CC interactions

Stability

David IETF95 Buenos Aires mdash ICCRG 6 9

Recursive Congestion SignalsISP 1

AccessISP 2

Access

GWGW

Public Internet Layer

Inter ISP Layer

ISP 2 Layer

ServerHost

ISP 1 LayerX

Feedback options in RINA

ImplicitI Push back and push up

ExplicitI Packet marking and

reflectionI Control packet notifications

Potentially provides

Faster responses

Smaller buffer requirements

Cautions mdash Current research

CC interactions

Stability

David IETF95 Buenos Aires mdash ICCRG 6 9

Recursive Congestion SignalsISP 1

AccessISP 2

Access

GWGW

Public Internet Layer

Inter ISP Layer

ISP 1 Layer ISP 2 Layer

ServerHost

X

Feedback options in RINA

ImplicitI Push back and push up

ExplicitI Packet marking and

reflectionI Control packet notifications

Potentially provides

Faster responses

Smaller buffer requirements

Cautions mdash Current research

CC interactions

Stability

David IETF95 Buenos Aires mdash ICCRG 6 9

Recursive Congestion SignalsISP 1

AccessISP 2

Access

GWGW

Public Internet Layer

Inter ISP Layer

ISP 1 Layer ISP 2 Layer

ServerHost

X

Feedback options in RINA

ImplicitI Push back and push up

ExplicitI Packet marking and

reflectionI Control packet notifications

Potentially provides

Faster responses

Smaller buffer requirements

Cautions mdash Current research

CC interactions

StabilityDavid IETF95 Buenos Aires mdash ICCRG 6 9

Stable CC inspired by Logistic Growth in Nature

0 2 4 6 8

K

t

Pop

ula

tion

P

(t)

P(t) = KP0ert

K+P0(ertminus1)

P-F Verhulst Notice sur la loi que la population poursuit dans sonaccroissement

Correspondance mathematique et physique 10113ndash121 1838

URL httpbooksgooglecomid=8GsEAAAAYAAJ

David IETF95 Buenos Aires mdash ICCRG 7 9

Congestion Control

Types of ldquosolvablerdquo Network Congestion

Network adaption ndash traffic rerouted along less congested paths

Network restriction ndash admission control and policing

End system adaption ndash traffic sources adjust send rates

Recursive CC work in RINA

Stability of chained and stacked congestion controllers

Congestion signals and how they influence performance

Stable scalable new types of congestion control

David IETF95 Buenos Aires mdash ICCRG 5 9

Recursive Congestion SignalsISP 1

AccessISP 2

Access

GWGW

Public Internet Layer

Inter ISP Layer

ISP 1 Layer ISP 2 Layer

ServerHost

Feedback options in RINA

ImplicitI Push back and push up

ExplicitI Packet marking and

reflectionI Control packet notifications

Potentially provides

Faster responses

Smaller buffer requirements

Cautions mdash Current research

CC interactions

Stability

David IETF95 Buenos Aires mdash ICCRG 6 9

Recursive Congestion SignalsISP 1

AccessISP 2

Access

GWGW

Public Internet Layer

Inter ISP Layer

ISP 2 Layer

ServerHost

ISP 1 LayerX

Feedback options in RINA

ImplicitI Push back and push up

ExplicitI Packet marking and

reflectionI Control packet notifications

Potentially provides

Faster responses

Smaller buffer requirements

Cautions mdash Current research

CC interactions

Stability

David IETF95 Buenos Aires mdash ICCRG 6 9

Recursive Congestion SignalsISP 1

AccessISP 2

Access

GWGW

Public Internet Layer

Inter ISP Layer

ISP 1 Layer ISP 2 Layer

ServerHost

X

Feedback options in RINA

ImplicitI Push back and push up

ExplicitI Packet marking and

reflectionI Control packet notifications

Potentially provides

Faster responses

Smaller buffer requirements

Cautions mdash Current research

CC interactions

Stability

David IETF95 Buenos Aires mdash ICCRG 6 9

Recursive Congestion SignalsISP 1

AccessISP 2

Access

GWGW

Public Internet Layer

Inter ISP Layer

ISP 1 Layer ISP 2 Layer

ServerHost

X

Feedback options in RINA

ImplicitI Push back and push up

ExplicitI Packet marking and

reflectionI Control packet notifications

Potentially provides

Faster responses

Smaller buffer requirements

Cautions mdash Current research

CC interactions

StabilityDavid IETF95 Buenos Aires mdash ICCRG 6 9

Stable CC inspired by Logistic Growth in Nature

0 2 4 6 8

K

t

Pop

ula

tion

P

(t)

P(t) = KP0ert

K+P0(ertminus1)

P-F Verhulst Notice sur la loi que la population poursuit dans sonaccroissement

Correspondance mathematique et physique 10113ndash121 1838

URL httpbooksgooglecomid=8GsEAAAAYAAJ

David IETF95 Buenos Aires mdash ICCRG 7 9

Recursive Congestion SignalsISP 1

AccessISP 2

Access

GWGW

Public Internet Layer

Inter ISP Layer

ISP 1 Layer ISP 2 Layer

ServerHost

Feedback options in RINA

ImplicitI Push back and push up

ExplicitI Packet marking and

reflectionI Control packet notifications

Potentially provides

Faster responses

Smaller buffer requirements

Cautions mdash Current research

CC interactions

Stability

David IETF95 Buenos Aires mdash ICCRG 6 9

Recursive Congestion SignalsISP 1

AccessISP 2

Access

GWGW

Public Internet Layer

Inter ISP Layer

ISP 2 Layer

ServerHost

ISP 1 LayerX

Feedback options in RINA

ImplicitI Push back and push up

ExplicitI Packet marking and

reflectionI Control packet notifications

Potentially provides

Faster responses

Smaller buffer requirements

Cautions mdash Current research

CC interactions

Stability

David IETF95 Buenos Aires mdash ICCRG 6 9

Recursive Congestion SignalsISP 1

AccessISP 2

Access

GWGW

Public Internet Layer

Inter ISP Layer

ISP 1 Layer ISP 2 Layer

ServerHost

X

Feedback options in RINA

ImplicitI Push back and push up

ExplicitI Packet marking and

reflectionI Control packet notifications

Potentially provides

Faster responses

Smaller buffer requirements

Cautions mdash Current research

CC interactions

Stability

David IETF95 Buenos Aires mdash ICCRG 6 9

Recursive Congestion SignalsISP 1

AccessISP 2

Access

GWGW

Public Internet Layer

Inter ISP Layer

ISP 1 Layer ISP 2 Layer

ServerHost

X

Feedback options in RINA

ImplicitI Push back and push up

ExplicitI Packet marking and

reflectionI Control packet notifications

Potentially provides

Faster responses

Smaller buffer requirements

Cautions mdash Current research

CC interactions

StabilityDavid IETF95 Buenos Aires mdash ICCRG 6 9

Stable CC inspired by Logistic Growth in Nature

0 2 4 6 8

K

t

Pop

ula

tion

P

(t)

P(t) = KP0ert

K+P0(ertminus1)

P-F Verhulst Notice sur la loi que la population poursuit dans sonaccroissement

Correspondance mathematique et physique 10113ndash121 1838

URL httpbooksgooglecomid=8GsEAAAAYAAJ

David IETF95 Buenos Aires mdash ICCRG 7 9

Recursive Congestion SignalsISP 1

AccessISP 2

Access

GWGW

Public Internet Layer

Inter ISP Layer

ISP 2 Layer

ServerHost

ISP 1 LayerX

Feedback options in RINA

ImplicitI Push back and push up

ExplicitI Packet marking and

reflectionI Control packet notifications

Potentially provides

Faster responses

Smaller buffer requirements

Cautions mdash Current research

CC interactions

Stability

David IETF95 Buenos Aires mdash ICCRG 6 9

Recursive Congestion SignalsISP 1

AccessISP 2

Access

GWGW

Public Internet Layer

Inter ISP Layer

ISP 1 Layer ISP 2 Layer

ServerHost

X

Feedback options in RINA

ImplicitI Push back and push up

ExplicitI Packet marking and

reflectionI Control packet notifications

Potentially provides

Faster responses

Smaller buffer requirements

Cautions mdash Current research

CC interactions

Stability

David IETF95 Buenos Aires mdash ICCRG 6 9

Recursive Congestion SignalsISP 1

AccessISP 2

Access

GWGW

Public Internet Layer

Inter ISP Layer

ISP 1 Layer ISP 2 Layer

ServerHost

X

Feedback options in RINA

ImplicitI Push back and push up

ExplicitI Packet marking and

reflectionI Control packet notifications

Potentially provides

Faster responses

Smaller buffer requirements

Cautions mdash Current research

CC interactions

StabilityDavid IETF95 Buenos Aires mdash ICCRG 6 9

Stable CC inspired by Logistic Growth in Nature

0 2 4 6 8

K

t

Pop

ula

tion

P

(t)

P(t) = KP0ert

K+P0(ertminus1)

P-F Verhulst Notice sur la loi que la population poursuit dans sonaccroissement

Correspondance mathematique et physique 10113ndash121 1838

URL httpbooksgooglecomid=8GsEAAAAYAAJ

David IETF95 Buenos Aires mdash ICCRG 7 9

Recursive Congestion SignalsISP 1

AccessISP 2

Access

GWGW

Public Internet Layer

Inter ISP Layer

ISP 1 Layer ISP 2 Layer

ServerHost

X

Feedback options in RINA

ImplicitI Push back and push up

ExplicitI Packet marking and

reflectionI Control packet notifications

Potentially provides

Faster responses

Smaller buffer requirements

Cautions mdash Current research

CC interactions

Stability

David IETF95 Buenos Aires mdash ICCRG 6 9

Recursive Congestion SignalsISP 1

AccessISP 2

Access

GWGW

Public Internet Layer

Inter ISP Layer

ISP 1 Layer ISP 2 Layer

ServerHost

X

Feedback options in RINA

ImplicitI Push back and push up

ExplicitI Packet marking and

reflectionI Control packet notifications

Potentially provides

Faster responses

Smaller buffer requirements

Cautions mdash Current research

CC interactions

StabilityDavid IETF95 Buenos Aires mdash ICCRG 6 9

Stable CC inspired by Logistic Growth in Nature

0 2 4 6 8

K

t

Pop

ula

tion

P

(t)

P(t) = KP0ert

K+P0(ertminus1)

P-F Verhulst Notice sur la loi que la population poursuit dans sonaccroissement

Correspondance mathematique et physique 10113ndash121 1838

URL httpbooksgooglecomid=8GsEAAAAYAAJ

David IETF95 Buenos Aires mdash ICCRG 7 9

Recursive Congestion SignalsISP 1

AccessISP 2

Access

GWGW

Public Internet Layer

Inter ISP Layer

ISP 1 Layer ISP 2 Layer

ServerHost

X

Feedback options in RINA

ImplicitI Push back and push up

ExplicitI Packet marking and

reflectionI Control packet notifications

Potentially provides

Faster responses

Smaller buffer requirements

Cautions mdash Current research

CC interactions

StabilityDavid IETF95 Buenos Aires mdash ICCRG 6 9

Stable CC inspired by Logistic Growth in Nature

0 2 4 6 8

K

t

Pop

ula

tion

P

(t)

P(t) = KP0ert

K+P0(ertminus1)

P-F Verhulst Notice sur la loi que la population poursuit dans sonaccroissement

Correspondance mathematique et physique 10113ndash121 1838

URL httpbooksgooglecomid=8GsEAAAAYAAJ

David IETF95 Buenos Aires mdash ICCRG 7 9

Stable CC inspired by Logistic Growth in Nature

0 2 4 6 8

K

t

Pop

ula

tion

P

(t)

P(t) = KP0ert

K+P0(ertminus1)

P-F Verhulst Notice sur la loi que la population poursuit dans sonaccroissement

Correspondance mathematique et physique 10113ndash121 1838

URL httpbooksgooglecomid=8GsEAAAAYAAJ

David IETF95 Buenos Aires mdash ICCRG 7 9

Logistic Growth Based Congestion Control

Competition for resources

incrementally for senders

xi (t + 1) = xi (t)ri

C minus xi (t) minus N minus 1

N

Nsumj=1

xj

+ xi (t)

Recursive layered congestion control ndash ongoing

predatorndashprey food chain based CC

has calculable convergence and stability properties

David IETF95 Buenos Aires mdash ICCRG 8 9

Conclusions and Future work

Stability and CC interactions also relate to the current Internet

New CC can work in the Internet as wellUpcoming publication

I P Teymoori M Welzl G Stein E Grasa R Riggio K Rausch and D Siracuss Congestion control in therecursive internetwork architecture (RINA)In IEEE International Conference on Communications (ICC) Next Generation Networking and InternetSymposium May 2016

to appear

I RINA congestion control in data centersI Congestion control signals and stability

OCARINA (Optimizations to Compel Adoption of RINA)(see httpstitanuiononode1403)

Acknowledgements

The authors are funded by the European Community under its Seventh FrameworkProgramme through the PRISTINE project (CNECT-ICT-619305) The views expressedare solely those of the authorsFor more information about PRISTINE see httpict-pristineeu

David IETF95 Buenos Aires mdash ICCRG 9 9

Conclusions and Future work

Stability and CC interactions also relate to the current Internet

New CC can work in the Internet as wellUpcoming publication

I P Teymoori M Welzl G Stein E Grasa R Riggio K Rausch and D Siracuss Congestion control in therecursive internetwork architecture (RINA)In IEEE International Conference on Communications (ICC) Next Generation Networking and InternetSymposium May 2016

to appear

I RINA congestion control in data centersI Congestion control signals and stability

OCARINA (Optimizations to Compel Adoption of RINA)(see httpstitanuiononode1403)

Acknowledgements

The authors are funded by the European Community under its Seventh FrameworkProgramme through the PRISTINE project (CNECT-ICT-619305) The views expressedare solely those of the authorsFor more information about PRISTINE see httpict-pristineeu

David IETF95 Buenos Aires mdash ICCRG 9 9

Extra slides

David IETF95 Buenos Aires mdash ICCRG 1 1

• Appendix