3rd Workshop on Advances in Slicing for Softwarized Infrastructures (S4SI 2020)
Panel: Network Slicing is multifaceted but does its approach and understanding need to be fragmented?
Abstract: Network Slicing keeps growing in significance in the academic and industrial communities. Network Slicing can be defined from different functional or behavioral perspectives, as well as from different viewpoints depending on the stakeholder (e.g., verticals, solution providers, infrastructure owners) and the technical domain (e.g. cloud data centers, radio access, packet/optical transport networks). Standardization bodies and open source projects are being involved in some forms of network slicing support. How far are these views from each other? Is fragmentation leading to incompatible approaches or is there some hope of convergence, at least at conceptual levels? What is the next frontier in Network Slicing? These and other questions will be thrown to our panel experts after introducing their lightning viewpoints.
Moderator: Christian Esteve Rothenberg, University of Campinas, Brazil
Panel Members
Constantine Polychronopoulos, Juniper Networks, USA
Uma Chunduri, Futurewei, USA
Slawomir Kuklinski, Orange Poland and Warsaw University of Technology, Poland
Stuart Clayman, University College London, UK
Augusto Venancio Neto, Federal University of Rio Grande do Norte, Brazil
Boost Fertility New Invention Ups Success Rates.pdf
Netsoft 2020 S4SI Workshop Panel
1. 3rd Workshop on Advances in Slicing for
Softwarized Infrastructures (S4SI 2020).
Panel
Monday June 29, 2020 – Full Day
4:15 p.m. – 5:15 p.m.
https://netsoft2020.ieee-netsoft.org/program-s4si-2020/ http://www.h2020-necos.eu/ - Novel Enablers for Cloud Slicing
5. Panel theme
Network Slicing is multifaceted
but does its approach and understanding need to be fragmented?
● Network Slicing keeps growing in significance in academia and industry
● Network Slicing can be defined from different functional or behavioral perspectives, as
well as from different viewpoints depending on the stakeholder (e.g., verticals, solution
providers, infrastructure owners) and the technical domain (e.g. cloud data centers, radio
access, packet/optical transport networks).
● Standardization bodies and open source projects are being involved in some forms of
network slicing support.
● How far are these views from each other?
● Is fragmentation leading to incompatible approaches or is there some hope of
convergence, at least at conceptual levels?
● What is the next frontier in Network Slicing?
S4SI 2020
8. 8
Public
CloudInternet
Centralized Control
(NS, Orchestration,
Service Provisioning)
Hard latency (and bandwidth) requirements
Distribution of content (CDNs) to edge
Local security & analytics driven ML/AL
Centralized control and RT analytics
Improved performance and UX
COTS hardware (OCP)
Latency
5-10ms
Managed Traffic
Breakout Traffic
Leveraging 1000s of edge clouds and public cloud providers
NS in a highly distributed multi-cloud environment
9. VP 5G & Telco Cloud
The Disruptive Potential of E2E
Network Slicing
Constantine Polychronopoulos
10. 10
Need is the mother of invention
Network Slicing and the 4th
Industrial Revolution
Enhanced Mobile Broadband
Gigabytes in a second
3D Video – 4K screens
Work & play in the cloud
Augmented reality
Industrial & vehicular automation
Mission critical broadband
Self Driving Car
Smart city cameras
Voice
Sensor NW
Low Latency
Ultra-high Reliability &
Low Latency
Massive loT
Massive Connectivity
12. 12
Unifying all access networks into 5G services
NFVI
vCU
vEPC
DU
vCU vCU
OLTONT
Switching Fabric
Residential
Services
Commercial
Services
Mobile Services
Intelligent RAN
Controller
Internet
ONT
ONT
vRAN & the RIC: Intelligent Edge Services and MEC
16. 16
NS 1
Cloudlet Cloudlet
Cloudlet
Cloudlet
Cloudlet
Cloudlet
Cloudlet
Cloudlet
Cloudlet
NS 2
Cloudlet Cloudlet
Cloudlet
NS n
Cloudlet Cloudlet
Cloudlet
Cloudlet
Cloudlet
Cloudlet
Cloudlet
Cloudlet
Cloudlet
Each NS is a subset of the
physical network
Cloudlets can be on prem,
managed or on CSPs
Multitenancy: Each NS has own
management, OSS/BSS &
analytics
Segmentation and customized
security policies
Service Provider: Maintains
global management of NSs, SLAs
aggregate capacity management,
scaling policies, accounting etc.
Leveraging 1000s of edge clouds and public cloud providers
NS in a highly distributed multi-cloud environment
17. 17
NS and ML/AI will be the cornerstone of the 4th
Industrial Revolution, redefining our entire industry
NS will Disrupt and Simplify cloud computing & networking
5G
Network
Slicing
Network
Fabric
Slicing
Tiered
Public
Clouds
4G LTE
Client
app
perf&
security
E2E
Network
Slicing
Tiered
SD-WAN
30. Sławomir Kukliński
● The NGMN White Paper about Network Slicing widely accepted (3GPP, ETSI
MANO, ...)
● In 5G all necessary mechanisms for slicing have been defined (selection, LCM,
run-time management, management delegation,...)
● GSMA has defined GST (Generic Slice Template) and NEST (Network Slice Type)
per use case
● Network Slice is a software entity - do we need to standardize everything?
● Still no large scale deployment of the technology
○ Vendors/verticals/operators are not ready yet?
○ Do we need to improve the existing approaches (RAN,...)?
○ Are there serious implementation issues? Do we have COTS solutions?
○ No market yet for network slicing?
○ Do we have killer application?
32. Augusto Neto
Considerations on Wireless Network
Slicing
Associate Prof. DIMAp/UFRN
Permanent member of PPgSC/UFRN
Leader of the REGINA research group
PQ-2 CNPq
36. Wireless Network Slicing:
Considerations
• Need to advance to spectrum-level towards full e2e slicing
capabilities
• SDR as enabling technology to enforce isolated service delivery at
the spectrum level
• AI-empowered wireless:
• ML can learn to predict settings best-fitted to the elasticity
• Predict demand and optimize channel settings
• Predict the load of each BS on the fly, figure out the location of devices
and decide about best BS to handover
40. Network Slicing is multifaceted but does its approach and understanding need to be
fragmented?
It doesn’t need to be, but it is, as it depends on where you start, and what is meant by
slicing ? It depends on your answer of what you think a slice is, but I believe it should be
something new.
A slice should be an addressable and manageable end-to-end abstraction, composed
from slice parts created over the underlying resources, but that means all the parts:
network, data center, storage, mobile edge, RAN, ...
41. Network Slicing is multifaceted but does its approach and understanding need to be
fragmented?
For slicing to be globally useful, in general, there needs to be agreement on:
● the right abstractions,
● the right building blocks,
● the right separation of concerns,
● the right composability,
● the right mechanisms to be in place for slice lifecycle management: to create, adjust,
and remove slices
42. How far are these views from each other?
Due to the different layering of the systems, and the different starting points we are
seeing some divergence.
It depends on how you construct a slice, which layer does the slicing, and what the
conceptual abstraction looks like.
Until you have decided these, you don't know what or how you can manage and
orchestrate them.
43. How far are these views from each other?
Some approaches suggested for slicing are not from first principles.
There are slicing processes which seem to rely on some magic. The mechanism to make
it work is left to someone else.
Other approaches make some big assumptions or rely on existing components which will
need the necessary extensions.
44. Is fragmentation leading to incompatible approaches or is there some hope of
convergence, at least at conceptual levels?
It is generally agreed that slicing gives an extra mechanism overlayed on the underlying
resources for the deployment of services with particular requirements.
We see standards being created by different bodies, with no testing or evaluation.
Renaming or relabelling an existing features and/or functions is not a slice:
A VPN is still a VPN ! You haven't created something new.
45. Is fragmentation leading to incompatible approaches or is there some hope of
convergence, at least at conceptual levels?
There is a naming anomaly where calling a slice "a network slice", but having VMs running
NFVs which are in a Data Center causes a bit confusing. If it uses DCs, then it's not just
the network part.
Having a non-symmetric model where you can partition a network, and partition the
storage, but not partition the Data Center gives poor composability, poor manageability
capabilities and characteristics.
Until the standards address these aspects, I think fragmentation will continue.
46. What is the next frontier in Network Slicing?
Concentrate on the aspects needed to complete:
● dynamic allocation of all slice parts
● end-to-end composability of the slice parts
● management interfaces for all slice parts
● create the lifecycle operations for slices
● building the right abstractions and layers for slicing
● design and build all new components and sub-systems need to make slicing work
● design APIs for addressing slices
47. What is the next frontier in Network Slicing?
Convergence of the concepts and the operations, and of course the standards.
The ability for operators to allocate network segments/parts, which meet slice
requirements, in a dynamic on-the-fly manner by returning all the end-point information,
in software, at run-time, and not in a written contract.
WIM on-demand – A modular approach for managing network slices
UCL + Telefonica
Thursday, July 2, 2020 PS6 – Slicing and Orchestration 1:30 p.m. – 3:00 p.m.: Virtual Room 1