5G as a service
The service function chaining
Davide Polonio & Marco Zanella December 2018
Outline
1 Introduction 2 5G
3 Architectural proposal 4 Proposal
5 Implementation 6 Tests
The current Internet situation
Increasing challenges
Need for more bandwidth
New verticals approaching
The current Internet situation
Increasing challenges
Need for more bandwidth New verticals approaching
5G - Improvement & design principles
5G design principles
Leverage spectrum
Cost-effective dense deployments Het-net: widespread coverage Simplify operations and
management
Common composable core Studied 5G enablers
SDN NFV
5G - Improvement & design principles
5G design principles Leverage spectrum
Cost-effective dense deployments Het-net: widespread coverage Simplify operations and
management
Common composable core Studied 5G enablers
SDN NFV
5G - Improvement & design principles
5G design principles Leverage spectrum
Cost-effective dense deployments Het-net: widespread coverage
Simplify operations and management
Common composable core Studied 5G enablers
SDN NFV
5G - Improvement & design principles
5G design principles Leverage spectrum
Cost-effective dense deployments Het-net: widespread coverage Simplify operations and
management
Common composable core Studied 5G enablers
SDN NFV
5G - Improvement & design principles
5G design principles Leverage spectrum
Cost-effective dense deployments Het-net: widespread coverage Simplify operations and
management
Common composable core
Studied 5G enablers
SDN NFV
5G - Improvement & design principles
5G design principles Leverage spectrum
Cost-effective dense deployments Het-net: widespread coverage Simplify operations and
management
Common composable core Studied 5G enablers
SDN NFV
5G - Improvement & design principles
5G design principles Leverage spectrum
Cost-effective dense deployments Het-net: widespread coverage Simplify operations and
management
Common composable core Studied 5G enablers
SDN
NFV
5G - Improvement & design principles
5G design principles Leverage spectrum
Cost-effective dense deployments Het-net: widespread coverage Simplify operations and
management
Common composable core Studied 5G enablers
SDN
5G - Network Function & SFC
Network Functions can perform
Packet classification Deep Packet Inspection Firewalling
Header enrichment Compression
Traffic Layer optimization . . .
A composable, ordered sequence of NFs forms a SFC
5G - Network Function & SFC
Network Functions can perform Packet classification
Deep Packet Inspection Firewalling
Header enrichment Compression
Traffic Layer optimization . . .
A composable, ordered sequence of NFs forms a SFC
5G - Network Function & SFC
Network Functions can perform Packet classification
Deep Packet Inspection
Firewalling
Header enrichment Compression
Traffic Layer optimization . . .
A composable, ordered sequence of NFs forms a SFC
5G - Network Function & SFC
Network Functions can perform Packet classification
Deep Packet Inspection Firewalling
Header enrichment Compression
Traffic Layer optimization . . .
A composable, ordered sequence of NFs forms a SFC
5G - Network Function & SFC
Network Functions can perform Packet classification
Deep Packet Inspection Firewalling
Header enrichment
Compression
Traffic Layer optimization . . .
A composable, ordered sequence of NFs forms a SFC
5G - Network Function & SFC
Network Functions can perform Packet classification
Deep Packet Inspection Firewalling
Header enrichment Compression
Traffic Layer optimization . . .
A composable, ordered sequence of NFs forms a SFC
5G - Network Function & SFC
Network Functions can perform Packet classification
Deep Packet Inspection Firewalling
Header enrichment Compression
Traffic Layer optimization
. . .
A composable, ordered sequence of NFs forms a SFC
5G - Network Function & SFC
Network Functions can perform Packet classification
Deep Packet Inspection Firewalling
Header enrichment Compression
Traffic Layer optimization . . .
A composable, ordered sequence of NFs forms a SFC
5G - Network Function & SFC
Network Functions can perform Packet classification
Deep Packet Inspection Firewalling
Header enrichment Compression
Traffic Layer optimization . . .
A composable, ordered sequence of NFs forms a SFC
5G - Network Function & SFC
Network Functions can perform Packet classification
Deep Packet Inspection Firewalling
Header enrichment Compression
Traffic Layer optimization . . .
A composable, ordered sequence of NFs forms a SFC
5G - Virtual Network Function (VNF)
Why do not we go “virtual”?
Benefits
Less maintenance costs Less time to deploy More flexibility Automation
5G - Virtual Network Function (VNF)
Why do not we go “virtual”?
Benefits
Less maintenance costs Less time to deploy More flexibility Automation
5G - Virtual Network Function (VNF)
Why do not we go “virtual”?
Benefits
Less maintenance costs
Less time to deploy More flexibility Automation
5G - Virtual Network Function (VNF)
Why do not we go “virtual”?
Benefits
Less maintenance costs Less time to deploy
More flexibility Automation
5G - Virtual Network Function (VNF)
Why do not we go “virtual”?
Benefits
Less maintenance costs Less time to deploy More flexibility
Automation
5G - Virtual Network Function (VNF)
Why do not we go “virtual”?
Benefits
Less maintenance costs Less time to deploy More flexibility Automation
5G - Network Slice
What?
Why?
5G - Network Slice
What?
Why?
5G - Network Slice
What?
Why?
5G - Network Slice Example
Section studied
Orchestrators
Virtual network functions (VNFs) can be orchestrated
Several orchestrators are available
Act as Virtual Infrastructure Manager (VIM)
Orchestrators
Virtual network functions (VNFs) can be orchestrated
Several orchestrators are available
Act as Virtual Infrastructure Manager (VIM)
Orchestrators
Virtual network functions (VNFs) can be orchestrated
Several orchestrators are available
Act as Virtual Infrastructure Manager (VIM)
Orchestrators
Virtual network functions (VNFs) can be orchestrated
Several orchestrators are available
Act as Virtual Infrastructure Manager (VIM)
Orchestrators
Virtual network functions (VNFs) can be orchestrated
Several orchestrators are available
Act as Virtual Infrastructure Manager (VIM)
Orchestrators
Virtual network functions (VNFs) can be orchestrated
Several orchestrators are available
Act as Virtual Infrastructure Manager (VIM)
Orchestrators
Virtual network functions (VNFs) can be orchestrated
Several orchestrators are available
Act as Virtual Infrastructure Manager (VIM)
MANagement and Orchestration
Tasks
Deploy VNF and orchestrate Service Function Chains (SFC)
Manage VNFs lifecycle Contains data repositories
Can handle one or multiple Point-of-Presence (PoP)
MANagement and Orchestration
Tasks
Deploy VNF and orchestrate Service Function Chains (SFC) Manage VNFs lifecycle
Contains data repositories
Can handle one or multiple Point-of-Presence (PoP)
MANagement and Orchestration
Tasks
Deploy VNF and orchestrate Service Function Chains (SFC) Manage VNFs lifecycle
Contains data repositories
Can handle one or multiple Point-of-Presence (PoP)
MANagement and Orchestration
Tasks
Deploy VNF and orchestrate Service Function Chains (SFC) Manage VNFs lifecycle
Contains data repositories
Can handle one or multiple Point-of-Presence (PoP)
Proposal - Our goal
Proposal - Final version
Final solution based on Harbor + Kubernetes
Use case example
VIBeS - ESA ARTES 6A.054
This thesis is inspired by European Space Agency VIBeS (VIrtualized NFs for Broadband Satellite networks) project The goal is to exploit NFV to enhance communication paths that involve a satellite link
Redesign Performance Enhancement Proxies (PEP) as virtualized functions
Current VNF/SFC implementations
The reference architecture of VNFs, SFCs and their management is ETSI NFV/MANO proposal
Most of the current VNF implementations use hypervisor-based virtualization
SFC deployed are statically defined at creation time
Our proposal
Exploit operating system virtualization (Docker) to deploy VNFs
Using Kubernetes as container orchestrator
Develop our manager and orchestrator (MANO) for the VNFs/SFCs and the other components to make possible the creation of chains
Testbed
Final proposal - High level view
Harbor: Management and Orchestration
Harbor is the MANO we developed
It allows to create definition of VNFs and SFCs and manages repositories on which they are stored
From definition is possible to run/stop chains
It is not deployed on Kubernetes but it can communicate with it
Ironhide: chains endpoints
Ironhide is the component that manages traffic that enters and exits the chain
It classify entering packets choosing the most suitable chain Classification based on transport layer protocol used Performed only at the edges of the chain, not during the traversal
Roulette: Context and Chains provider
Middleware between the SFC components and the MANO
Allows to access chain definitions through APIs
Preserve end-to-end information on chains’ endpoints used
Astaire: Service Function Forwarder
It is the component that logically create the path for traffic
Responsable for forward traffic to the VNFs and manage the response
It queries Roulette to check the next hop based on chain
Tests
Tests based on system responsiveness Docker vs VirtualBox startup
SFC length startup
Tests
Tests based on system responsiveness Docker vs VirtualBox startup SFC length startup
Tests - Docker vs VirtualBox
Tests - SFC length startup
Overhead due to Pod scheduling Resource allocation Route update
Average start up time (in seconds)
Tests - RTT
RTT is higher using the SFC platform
Tests - Time to recover from a fault
after 5s that the sender starts to send packet the VNF is stopped
about 13s required to restore a faulty container
Future work
Integrate Harbor with Openbaton and expand its features
Rudimental classifier must be refined adding more capabilities
Ironhide implementation requires lower level connectivity to be able to read the whole TCP and UDP headers
Conclusion
In this work we developed a proof of concept implementation of an SFC platform
Differently from other current implementation network function and chains are deployed using Kubernetes and Docker
Zbigniew Kotulski et al.“On end-to-end approach for slice isolation in 5G networks. Fundamental challenges”. In:
Computer Science and Information Systems (FedCSIS), 2017 Federated Conference on. IEEE. 2017, pp. 783–792.
NGMN Alliance.“Description of network slicing concept”.
In: NGMN 5G P 1 (2016).
Jose Ordonez-Lucena et al.“Network slicing for 5g with SDN/NFV: concepts, architectures and challenges”. In:
arXiv preprint arXiv:1703.04676 (2017).
Di Liu and Libin Zhao.“The research and implementation of cloud computing platform based on docker”.In: Wavelet Active Media Technology and Information Processing (ICCWAMTIP), 2014 11th International Computer Conference on. IEEE. 2014, pp. 475–478.
Michael Eder.“Hypervisor-vs. container-based
virtualization”.In: Future Internet (FI) and Innovative Internet Technologies and Mobile Communications (IITM) 1 (2016).
NM Mosharaf Kabir Chowdhury and Raouf Boutaba.
“Network virtualization: state of the art and research challenges”.In: IEEE Communications magazine 47.7 (2009).
Rashid Mijumbi et al.“Network function virtualization:
State-of-the-art and research challenges”.In: IEEE Communications Surveys & Tutorials 18.1 (2016), pp. 236–262.
Mikio Iwamura. “NGMN view on 5G architecture”.In:
Joel Halpern and Carlos Pignataro.Service function chaining (sfc) architecture.Tech. rep. 2015.
Ahmed M Medhat et al.“Service function chaining in next generation networks: State of the art and research
challenges”.In: IEEE Communications Magazine 55.2 (2017), pp. 216–223.
