A Fully-Flexible Internet Architecture for the Next-Generation - - PowerPoint PPT Presentation

▶
a fully flexible internet architecture
SMART_READER_LITE
LIVE PREVIEW

A Fully-Flexible Internet Architecture for the Next-Generation - - PowerPoint PPT Presentation

Jul 03, 2023 241 likes •833 views

2 nd Visions for Future Communications Summit Technologies and Services Towards 6G Introducing FlexNGIA: A Fully-Flexible Internet Architecture for the Next-Generation Tactile Internet Mohamed Faten Zhani cole de technologie suprieure

slide-1
SLIDE 1

Introducing FlexNGIA: A Fully-Flexible Internet Architecture for the Next-Generation Tactile Internet

Lisbon, Portugal, November 28, 2019

Mohamed Faten Zhani

École de technologie supérieure (ÉTS Montreal) Canada

2nd Visions for Future Communications Summit Technologies and Services Towards 6G

slide-2
SLIDE 2

Outline

  • A Glance into the Future
  • Limitations of Today’s Internet
  • FlexNGIA: Fully-Flexible Next-Generation Internet Architecture
  • Use Cases
  • Research Challenges

2

  • Keynote at ACM SIGCOMM 2019 Workshop on Networking for Emerging Applications and Technologies (NEAT 2019)
  • M. F. Zhani, H. ElBakoury, “FlexNGIA: A Flexible Internet Architecture for the Next-Generation Tactile Internet,”

ArXiV 1905.07137, May 17, 2019 https://arxiv.org/abs/1905.07137

slide-3
SLIDE 3

A Glance into the Future

Future Applications

  • Telepresence
  • Virtual Reality
  • Augmented Reality
  • Holoportation
  • Haptics
  • …

3

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)
slide-4
SLIDE 4

Loading…

4

slide-5
SLIDE 5

Welcome to the Matrix

5

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)
slide-6
SLIDE 6

Future Applications Requirements & Characteristics

  • Characteristics
  • Octopus-like applications: huge number of flows

for each application

  • Changing requirements : requirements can change
  • ver time
  • Requirements:
  • High processing power: real-time processing
  • High bandwidth (e.g., VR (16K, 240 fps)  31.85 Gbps)
  • Ultra-low Latency: 1ms to 20ms
  • Multi-flow synchronization
  • High availability

6

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)
slide-7
SLIDE 7

Outline

  • A Glance into the Future
  • Limitations of Today’s Internet
  • FlexNGIA: Fully-Flexible Next-Generation Internet

Architecture

  • Use cases
  • Research Challenges

7

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)
slide-8
SLIDE 8

Today’s Internet Limitations

  • A network of networks
  • No control over E2E performance
  • Offered service: “Best effort” data delivery.. no more
  • Transport Layer Protocols
  • One-size-fits-all service offering: TCP offers reliability, packet

retransmission, congestion and flow control

  • Blind congestion control
  • The two end points limitation
  • Same limitations are inherited by QUIC and SCTP

ISP2 ISP3 ISP1 ISP4

8

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)
slide-9
SLIDE 9

Today’s Internet Limitations (2)

  • Network layer protocols and services
  • Not aware of the applications characteristics and requirements

(which flow belongs to which app?, priorities?)

  • No collaboration with upper layers (transport, application)
  • The network knows better about the congestion (e.g., location, severity)
  • The network could help with reliability (e.g., retransmission)…
  • Packet header
  • Not easy to include metadata or commands
  • Whose is going to use such metadata/commands?

9

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)
slide-10
SLIDE 10

Outline

  • A Glance into the Future
  • Limitations of Today’s Internet
  • FlexNGIA: Fully-Flexible Next-Generation Internet

Architecture

  • Use cases
  • Research Challenges

10

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)
slide-11
SLIDE 11

FlexNGIA

Computing resources Application-Aware Network Services Business model Flexible headers

  • In-Network

Computing: any function anywhere

  • Advanced functions

tailored to applications

  • App-aware traffic

engineering

FlexNGIA

  • Network

providers offer Service Function Chains/meshes

  • Stringent

performance requirements

  • Tailored

to the application

Cross-layer Design (Transport+Network)

  • Breaking the

end-to-end paradigm

  • In-network advanced

transport functions

  • Better congestion control
  • Stringent performance

and reliability guarantees

11

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)
slide-12
SLIDE 12

Future Internet Services

Service Function Chain/Mesh (SFC):

  • Multiple sources and destinations
  • Define your Network Functions: type, software,

behavior, input/output packet format, expected processing delay, buffer size

  • Define communication protocols

(layer 3 and above)

  • Define performance requirements

(e.g., throughput, packet loss, end-to-end delay, jitter)

12

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)

S1

1

2 6 7 5 9 12 8 10

POP Link Virtual Link

S0 S2 D=13

D S

Source Destination

Service Function Chain Physical Infrastructure

IDS

Firewall

NAT

Mapping

Mapped instance

3 4 13 11

slide-13
SLIDE 13

Network Protocol Stack/Functions

  • Basic Network Functions

(e.g., packet forwarding)

  • Advanced Network Functions:
  • Could operate at any layer
  • Only limited by our imagination

13 Transport Application Network Link

Application Assistant OSAP Application-aware Network Functions Monitoring & Measurement Network Functions Transport Assistant Transport and Traffic Engineering Network Functions

  • Examples: packet grouping, caching and retransmission,

data processing (e.g., image/video cropping, compression, rendering, ML), application-aware flow multiplexing (e.g., incorporating/merging data) Functions could break the end-to-end principle SDN++: SDN should go beyond configuring forwarding rules and should provide the ability to dynamically configure these new functions

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)
slide-14
SLIDE 14

Outline

  • A Glance into the Future
  • Limitations of Today’s Internet
  • FlexNGIA: Fully-Flexible Next-Generation Internet Architecture
  • Use Cases
  • Research Challenges

14

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)
slide-15
SLIDE 15

Application-Aware Network Services

Application Assistant (AA)

  • One AA at each end-point
  • Interfaces with objects/sensors
  • Measures the application performance

and user QoE

  • Identifies the applications’ requirements

at run-time

  • Adds additional metadata to be used

by subsequent Network Functions  Application-Aware Network Services

15

S1 S0 S2 D0

NF12 NF11 AA AA AA AA Sensors/

  • bjects

D1

AA

D2

AA NF13

D S

Traffic Source Traffic Destination AA Application Assistant Sensors/objects

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)
slide-16
SLIDE 16

Network-Assisted Reliable Data Transport

  • Goal
  • Minimize retransmission delay
  • Improved congestion control
  • Solution: service chain with a "transport

Assistant" function

  • Services of the Transport Assistant:
  • Combines transport and network layers
  • Cache and retransmit packets
  • Detect packet loss
  • Routing and congestion control: adjusting rate,

dropping packets

16

1 2 6 7 5 9 12 8 10 S0 D=13

Network-Assisted Transport Physical Infrastructure

Transport Assistant

Mapping

3 4 13 11

POP Link Virtual Link

D S

Source Destination Mapped instance

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)
slide-17
SLIDE 17

Mixed Virtual Reality and Holograms

  • Users are exploring a virtual reality

environment with several human holograms and objects

  • Challenges
  • How many intermediate functions?
  • What kind of functions?
  • How the traffic should steered from the

flow sources?

  • How many instances for each function?
  • Where to place them?
  • Example of deployement
  • Encoder: encode and compress video
  • Transport manager: congestion control
  • Video cropper: crop 3D objects

17

H2

3 4 5 6 7 8 9 1 2 11

1 2 6 7 5 9 12 8 10 H1 VR D2 Application s Service Function Chain

Physical Infrastructure Transport Assistant Video Cropper

Mapping

3 4 13 11 Virtual Topology

Destinations Encoder

D3 D1

12 13 Sources Holograms + Virtual reality

AA AA AA AA AA AA

Point of Presence Physical Link Virtual Link

D S

Traffic Source Traffic Destination Mapped Instance AA Application Assistant Service Function Chain SFC Sensors/objects

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)
slide-18
SLIDE 18

Research Challenges

  • Designing Service Function Meshes tailored to applications
  • High-performance softwarized functions
  • Signaling
  • Slicing/Resource Allocation
  • Fault-tolerance and Failure Management
  • High-precision and fine-grained monitoring and measurements
  • SDN++
  • Distributed cross-layer transport protocol (socket, caching,

communication)

  • Security and Privacy
  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)

18

slide-19
SLIDE 19

Looking for More Details?

  • Mohamed Faten Zhani, https://profs.etsmtl.ca/mfzhani/
  • FlexNGIA Project: M. F. Zhani, H. ElBakoury, “FlexNGIA: A Flexible

Internet Architecture for the Next-Generation Tactile Internet,”ArXiV 1905.07137, May 17, 2019 https://arxiv.org/abs/1905.07137

  • Keynote at ACM SIGCOMM 2019 Workshop on Networking

for Emerging Applications and Technologies (NEAT 2019)

19

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)

École de technolgie supérieure (ÉTS Montreal) University of Quebec, Canada

slide-20
SLIDE 20

Thank You

Questions

20

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)
slide-21
SLIDE 21
  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)

21

slide-22
SLIDE 22
  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)

22

slide-23
SLIDE 23

FlexNGIA A Fully Flexible Novel Architecture for the Next-Generation Tactile Internet

Beijing, China, 19 August 2019

Mohamed Faten Zhani

École de technologie supérieure (ÉTS Montreal) University of Quebec Canada ACM SIGCOMM 2019 Workshop on Networking for Emerging Applications and Technologies (NEAT 2019)

slide-24
SLIDE 24

Outline

  • A Glance into the Future
  • Limitations of Today’s Internet
  • FlexNGIA: Fully-Flexible Next-Generation Internet Architecture
  • Use cases
  • Conclusion

24

  • M. F. Zhani, H. ElBakoury, “FlexNGIA: A Flexible Internet Architecture

for the Next-Generation Tactile Internet,” ArXiV 1905.07137, May 17, 2019 https://arxiv.org/abs/1905.07137

slide-25
SLIDE 25

A Glance into the Future

Future Applications

  • Telepresence
  • Virtual Reality
  • Augmented Reality
  • Holoportation
  • Haptics
  • …

25

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)
slide-26
SLIDE 26

Loading the Matrix…

26

slide-27
SLIDE 27

Welcome to the Matrix

27

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)
slide-28
SLIDE 28

Future Applications Requirements & Characteristics

  • Characteristics
  • Octopus-like applications: huge number of flows

for each application

  • Changing requirements : requirements can change
  • ver time
  • Requirements:
  • High processing power: real-time processing
  • High bandwidth (e.g., VR (16K, 240 fps)  31.85 Gbps)
  • Ultra-low Latency: 1ms to 20ms
  • Multi-flow synchronization
  • High availability

28

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)
slide-29
SLIDE 29

Outline

  • A Glance into the Future
  • Limitations of Today’s Internet
  • Internet Infrastructure and Services
  • Network Stack Layers and Headers
  • FlexNGIA: Fully-Flexible Next-Generation Internet Architecture
  • Use cases
  • Conclusion

29

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)
slide-30
SLIDE 30

Internet Infrastructure and Services

  • A network of networks
  • Offered service: “Best effort”

data delivery.. no more

  • No control over the infrastructure

No control over the end-to-end path and quality of service No performance guarantees

ISP2 ISP3 ISP1 ISP4

30

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)
slide-31
SLIDE 31

Transport Layer Protocols

Many modern protocols like SCTP and QUIC but let’s focus first on TCP..

  • One-size-fits-all service offering: TCP offers reliability, data

retransmission, congestion and flow control

  • Blind Congestion control
  • The two end points limitation:
  • High retransmission delays (~ 3x e2e delay)
  • Transport and network layers are not aware which flows belong to the same

application

  • M. F. Zhani - FlexNGIA 2019

31

slide-32
SLIDE 32

Network Layer Protocols

  • Not aware of the applications
  • The application composition (in terms of flows)
  • Performance requirements of each of these flows and how these

requirement change over time Drop packets « blindly »

  • No collaboration with the transport layer
  • Do not provide explicit feedback or support to transport layer

(maybe ECN is interesting but it is not enough)

  • Do not help with other transport services (e.g., reliability)

32

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)
slide-33
SLIDE 33

Network Stack header

Problems with current headers:

  • Do not provide additional informations about objects/sensors, flows

belonging to the same application, applications’ requirements, etc.

  • Not flexible enough: It is not easy to incorporate meta-data

and commands

33

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)
slide-34
SLIDE 34

Outline

  • A Glance into the Future
  • Limitations of Today’s Internet
  • FlexNGIA: Fully-Flexible Next-Generation Internet Architecture
  • Future Internet Infrastructure and services
  • Business Model
  • Management Framework
  • Network Protocol Stack/Functions
  • Stack Headers
  • Use cases
  • Conclusion

34

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)
slide-35
SLIDE 35

Future Internet Infrastructure and Services

How a network will look like?

  • Computing resources are everywhere:

Available at the edge and at the core

  • f the network
  • Commodity servers but also dedicated

hardware, FPGA, GPU, NPU, etc.

 In-Network computing Reduce steering delay Full Programmability: Any function could be provisioned anywhere (virtual machines/containers)

1

2 6 7 5 9 12 8 10 3 4 13 11

Cloud Data Center Micro cloud

35

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)
slide-36
SLIDE 36

Future Internet Infrastructure and Services

How does Future Internet look like?

  • Still a network of networks..
  • What is new?
  • More services: Service Function chains

More advanced functions More than just delivery

  • Stringent performance guarantees

36

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)
slide-37
SLIDE 37

Future Internet Infrastructure and Services

Service Function Chain (SFC)

  • Multiple connected network functions
  • Multiple sources and destinations
  • Made out from Network Functions
  • Defines, for each network function, the type,

software, input/output packet format, expected processing delay, buffer size

  • Defines performance requirements

(e.g., throughput, packet loss, end-to-end delay, jitter)

37

S1 S0 S2 D0

NF12 NF11 AA AA AA AA Sensors/

  • bjects

D1

AA

D2

AA NF13

D S

Traffic Source Traffic Destination AA Application Assistant Sensors/objects

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)
slide-38
SLIDE 38

Business Model

Network Operators

  • Own and manage the physical

infrastructure (i.e., one network)

  • Deploy platforms and software

required to run network functions

  • The service could be simply data

delivery or a SFC

  • Provision and manage SFCs

S1

1

2 6 7 5 9 12 8 10

POP Link Virtual Link

S0 S2 D=13

D S

Source Destination

Service Function Chain Physical Infrastructure

IDS

Firewall

NAT

Mapping

Mapped instance

3 4 13 11

38

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)
slide-39
SLIDE 39

Business Model (cont)

Customers

  • Could be other network operators, companies or Institutions
  • Define the required SFC and Identify the chain sources/destinations
  • Rely on the operator to provision and manage the SFC and satisfy SLA
  • SFC composition
  • SLA requirements for the SFC
  • Bandwidth
  • End-to-end delay
  • Reliability, availability
  • SLA requirements for each NFs
  • Processing power
  • Packet format(s)
  • Packet drop criteria…

Customer

IDS

Firewall

Operator

NAT

39

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)
slide-40
SLIDE 40

Business Model (cont)

  • Example of potential Network Operators:
  • ISPs (e.g., AT&T or Bell Canada) and web-scale

companies (e.g., Google, Facebook, Amazon)

  • Example: Google Cloud Platform
  • World wide global Infrastructure
  • Software defined platform
  • Full control over the infrastructure

15 Data centers 100 Points of Presence (PoPs) 1000+ Edge nodes

Source: cloud.google.com 40

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)
slide-41
SLIDE 41

Resource Management Framework

Resource Allocation

  • The Service Function Chain (SFC)

is defined by the application designer

  • 2-step resource allocation:
  • Translation: the SFC is translated

into a virtual topology

  • Mapping: virtual topology

are mappa

41

3 4 5 6 7 8 9 10 1 2 11

1 2 6 7 5 9 12 8 10

Physical Infrastructure Mapping

3 4 13 11

Virtual Topology

S1 S0 S2 D13

Service Function Chain SFC1 associated with Application 1 NF12 NF11 NF13 AA AA AA AA AA AA AA AA AA AA AA AA Translation Sensors/

  • bjects

Point of Presence Physical Link Virtual Link

D S

Traffic Source Traffic Destination Mapped Instance AA Application Assistant Service Function Chain SFC Sensors/objects

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)
slide-42
SLIDE 42

Resource Management Framework

Main components:

  • Signaling module
  • Application Control Module
  • Ressource allocation Module

42

S1

1

2 6 7 5 9 12 8 10 S0 S2 D1

Application1 - SFC1 Physical Infrastructure NF12 NF11 NF13

Mapping

3 4 13 11

Resource Management Framework

Monitoring Module Application Control Module Monitoring Data Failure Management Module AA AA AA AA .. .. AA AA AA Resource Allocation Module NB: For simplicity, the figure shows only the mapping of the chain SFC1 associated to Application 1

S6 S5 S7 DN

Applicationn - SFCn NFn2 NFn1 NFn3 Monitoring Data AA AA AA AA Application Control Module Sensors/

  • bjects

Signaling Module Commands Monitoring Data Point of Presence Physical Link Virtual Link

D S

Traffic Source Traffic Destination Mapped Instance AA Application Assistant Service Function Chain SFC Sensors/objects AA

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)
slide-43
SLIDE 43

Network Protocol Stack/Functions

  • Basic Network Functions

(e.g., packet forwarding)

  • Advanced Network Functions:
  • Could operate at any layer
  • Only limited by our imagination

43 Transport Application Network Link

Application Assistant OSAP Application-aware Network Functions Monitoring & Measurement Network Functions Transport Assistant Transport and Traffic Engineering Network Functions

  • Examples: packet grouping, caching and retransmission,

data processing (e.g., image/video cropping, compression, rendering, ML), application-aware flow multiplexing (e.g., incorporating/merging data) Functions could break the end-to-end principle SDN++: SDN should go beyond configuring forwarding rules and should provide the ability to dynamically configure these new functions

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)
slide-44
SLIDE 44

Network Protocol Stack/Functions Application Assistant

Application Assistant (AA)

  • One AA at each end-point
  • Interfaces with objects/sensors
  • Measures the application performance

and user QoE

  • Identifies the applications’ requirements

at run-time

  • Adds additional metadata To be used

by subsequent Network Functions  Application-Aware Network Services

44

S1 S0 S2 D0

NF12 NF11 AA AA AA AA Sensors/

  • bjects

D1

AA

D2

AA NF13

D S

Traffic Source Traffic Destination AA Application Assistant Sensors/objects

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)
slide-45
SLIDE 45

Network Protocol Stack/Functions Transport Assistant

Transport Assistant (TA)

  • A cross-layer Network Function
  • Combines services of the transport

and network layers

  • Manages all the flows
  • f the same application
  • Implements Transport/Network

functions (e.g., congestion control,

packet loss detection, packet cache and retransmission, routing)

  • One or multiple TA could be

provisioned in the same SFC

45

Transport Layer (TCP) Application Layer Link Layer Cross-Layer Transport

  • E2E communication
  • Blind congestion Control
  • Inaccurate Packet Loss Detection
  • Guaranteed Reliability
  • E2E Packet Retransmission Process

Network Layer

  • IP protocol (header and addressing)
  • Routing Protocols/SDN
  • ICMP for Control Information
  • No Advanced Network Functions
  • Multi-point communication
  • Network-assisted congestion control
  • Network-assisted reliability and performance

guarantees

  • Accurate packet loss detection
  • Variable performance and reliability

Requirements over time

  • Variable Header
  • Meta-data and commands within packet

headers

  • Advanced Network Functions
  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)
slide-46
SLIDE 46

Network Stack Headers

  • Signaling packets
  • Instantiate an application
  • Convey application requirements
  • Data packets: carry data
  • Layer 2 header: contains mainly the application id used for packet

forwarding (similar to VLANs)

  • Upper layers:
  • Fully flexible header format (customizable meta-data and commands)
  • Defined depending on the application
  • Network functions should be aware of the expected format

46

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)
slide-47
SLIDE 47

Outline

  • A Glance into the Future
  • Limitations of Today’s Internet
  • FlexNGIA: Fully-Flexible Next-Generation Internet Architecture
  • Use Cases scenarios
  • Conclusion

47

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)
slide-48
SLIDE 48

Network-Assisted Data Transport

  • Goal
  • Minimize retransmission delay
  • Improved congestion control
  • Solution: service chain with a "transport

Assistant" function

  • Service of the Transport Assistant:
  • Caching and retransmissting packets
  • Detecting packet loss
  • Congestion control: adjusting rate, dropping

packets, compression

48

1 2 6 7 5 9 12 8 10 S0 D=13

Network-Assisted Transport Physical Infrastructure

Transport Assistant

Mapping

3 4 13 11

POP Link Virtual Link

D S

Source Destination Mapped instance

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)
slide-49
SLIDE 49

Mixed Virtual Reality and Holograms

  • Users are exploring a virtual reality

environment with several human holograms and objects

  • Challenges
  • How many intermediate functions?
  • What kind of functions?
  • How the traffic should steered from the

flow sources?

  • How many instances for each function?
  • Where to place them?
  • Example of deployement
  • Encoder: encode and compress video
  • Transport manager: congestion control
  • Video cropper: crop 3D objects

49

H2

3 4 5 6 7 8 9 1 2 11

1 2 6 7 5 9 12 8 10 H1 VR D2 Application s Service Function Chain

Physical Infrastructure Transport Assistant Video Cropper

Mapping

3 4 13 11 Virtual Topology

Destinations Encoder

D3 D1

12 13 Sources Holograms + Virtual reality

AA AA AA AA AA AA

Point of Presence Physical Link Virtual Link

D S

Traffic Source Traffic Destination Mapped Instance AA Application Assistant Service Function Chain SFC Sensors/objects

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)
slide-50
SLIDE 50

Outline

  • A Glance into the Future
  • Limitations of Today’s Internet
  • FlexNGIA: Fully-Flexible Next-Generation Internet

Architecture

  • Use Cases
  • Conclusion

50

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)
slide-51
SLIDE 51

Conclusion

Computing resources Application-Aware Network Management Business model Flexible headers

  • In-Network

Computing: any function anywhere

  • Advanced functions

tailored to applications

  • App-aware traffic

engineering

FlexNGIA

  • Multiple source

destination Service Function Chains

  • Stringent

performance requirements

  • Tailored

to the application

Cross-layer Design (Transport+Network)

  • Breaking the

end-to-end paradigm

  • In-network advanced

transport functions

  • Better congestion control
  • Stringent performance

and reliability guarantees

51

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)
slide-52
SLIDE 52

Looking for More Details?

  • Mohamed Faten Zhani, https://profs.etsmtl.ca/mfzhani/
  • FlexNGIA Project: M. F. Zhani, H. ElBakoury, “FlexNGIA: A Flexible

Internet Architecture for the Next-Generation Tactile Internet,”ArXiV 1905.07137, May 17, 2019 https://arxiv.org/abs/1905.07137

52

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)

École de technolgie supérieure (ÉTS Montreal) University of Quebec, Canada

slide-53
SLIDE 53

Thank You

Questions

53

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)
slide-54
SLIDE 54

Research Challenges

  • Designing Service Function Chains tailored to applications
  • High-performance softwarized functions
  • Signaling
  • Resource Allocation
  • Fault-tolerance and Failure Management
  • High-Precision and Fine-grained Monitoring and Measurements
  • SDN++
  • Distributed Cross-Layer Transport Protocol (sockets, caching, communication)
  • Security and Privacy

Details are available in the paper (https://arxiv.org/abs/1905.07137)

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)

54

slide-55
SLIDE 55

Transport Layer Protocols (cont)

55

QUIC

  • Transport over UDP
  • Multi-streaming:
  • Every stream is a reliable

bidirectional bytestream

  • Multiplexed streams between

two endpoints

  • Stream prioritization
  • Flow-control and congestion

control very similar to TCP

  • Endpoints use Explicit Congestion

Notification (ECN) SCTP

  • Basically, a TCP++
  • Multi-streaming
  • Unordered delivery is possible
  • Flow control and congestion

control similar to TCP

slide-56
SLIDE 56

Transport Layer Protocols (cont)

What are the limitations of SCTP and QUIC?

  • E2E communication: multiple flows (streams) of the same application

may connect more than two end-points

  • A blind congestion control
  • No support from the network: the network knows better about its state

Can better locate and manage congestion Predict and detect more efficiently congestions/failures/problems… Can retransmit faster Can provide better garantees in terms of delay and packet loss

56

slide-57
SLIDE 57

Network Protocol Stack/Functions Transport Assistant (cont)

  • Transport Assistants manage all these flows while taking into account

that they all belong to the same application

  • TAs monitor these flows, divide the total bandwidth allocated for the

application among them.

(a) TCP and UDP (c) FlexNGIA Physical end-host Transport end point Transport layer connection (b) SCTP and QUIC

Illustration of how one single network application might be seen at the transport Layer

57

  • M. F. Zhani, H. ElBakoury - FlexNGIA 2019 (https://arxiv.org/abs/1905.07137)