3GPP Network Slicing

Generic / Static Slicing & Dynamic Slicing

Various deployment scenarios

Slicing – Edge Stack 1

 Operator software - NFs, Apps,
Network Slicing Phase 1 – Starting Point for providers Mgmt

(Note: NF placement in various Edges is just one example)

All kinds of traffic flows are considered equal.

Traditional OSS/BSS Only QoS treatment is taken care based on
the traffic priority.
Network Service & Complex App Orchestrator
(NFVO for K8s such as EMCO ) Essentially, no slicing (or considered as one
slice).

Works good for generic traffic such as Voice

calls, Consumer broadband and SMS etc..).
RU

Does this work for other use cases that

require ultra low latency, higher security
RU isolation and support for low power devices
Apps
(to support Enterprise private networking
Apps dSMF Apps over 5G)? Answer is NO.

5GC
RU Apps CU-CP dAMF Control
Notes:
DU CU-UP dUPF cUPF At each location, there could be multiple instances of a given NF. 3GPP
Selection procedures will take care of selecting the right NF during PDU
Device Edge Towers session establishment.
‘d’ in dUPF, dSMF and dAMF indicating ‘Distributed’. Normally, they
Near Cell Regional are expected at central place. But, for scalability reasons, thinking is to
Telco CO Core DC
tower DC deploy at multiple places as these are mostly commonly used Control
Edges
Slicing – Edge Stack Edges Plance CNFs 2
Network Slicing Phase 1.5 – Enterprise Application Slices Operator software - NFs, Apps,
Mgmt
(aka Multi Access Edge computing)

Traditional OSS/BSS
Operators allow Enterprises to deploy their
Network Service & Complex App Orchestrator applications near dUPF and cUPF for
(NFVO for K8s such as OpenNESS-EMCO) distributed computing.

Already happening to some extent –

Operators partnering with Hyperscalers and
others to provide managed services to
RU Enterprises (Example: Verizon and AWS
wavelength).

Belief is that operators would themselves

RU
Apps provide application slices on the same
infrastructure as their NFs for better resource
Apps dSMF Apps
usage (Need good security and performance
Apps CU-CP dAMF
5GC isolation solutions though)
RU Control
Note: These a re typical use cases – Enterprise application a nd there is no special
need for ultra l ow l atency.
DU CU-UP dUPF cUPF
EMCO provi des this functionality to great extent (Traffic Steering is a roadmap
Device Edge Towers fea ture)

Near Cell Regional

Telco CO Core DC
tower DC
Edges
Slicing – Edge Stack Edges 3
Network Slicing Phase 1.5 – Enterprise Application Slices Operator software - NFs, Apps,
Mgmt
With Operators providing security for applications.

Traditional OSS/BSS Operators allow Enterprises to deploy their

applications near dUPF and cUPF for
Network Service & Complex App Orchestrator distributed computing.
(NFVO for K8s such as OpenNESS-EMCO)
Operators provide security services to the
Enterprises who deploy applications.

In this example: Enterprise 1 and Enterprise 2

RU security is provided by provider via two slices
(Red). Multiple security chains are provided
for Enterprise1 slices on the traffic between
UPF and Enterprise Apps. In case of
RU
Enterprise2, on chain of function is good
enough for both application slices.

RU
Note: These a re typical use cases – Enterprise application a nd there is no special
need for ultra l ow l atency.

EMCO provi des this functionality to great extent (Traffic Steering is a roadmap
Device Edge Towers fea ture)

Near Cell Regional

Telco CO Core DC
tower DC
Edges
Slicing – Edge Stack Edges 4
 Operator software - NFs, Apps,

Network Slicing Phase 2 – Multiple generic slices Mgmt

Traditional OSS/BSS Hypothesis:

Simpler Slicing Automation
• Support for 3 or few generic slices for
Network Service & Complex App Orchestrator different use cases.
(NFVO for K8s such as EMCO ) • No NF Sharing
• 1:1 core and RAN slices
• Common infrastructure
• NFs as CNFs.
• Simpler Slicing Orchestration as there is no
RU dynamic slice creation/deletion.
• All slices are provider controlled.
• Only provider applications in each slice.

RU Why?
- Simpler deployment
mMTC
- Good security isolation
- Easy to visualize and control resources as
RU eMBB
they share common infrastructure
- No major enhancements to NFs ( Need to
SNSSAI aware though)
Device Edge Towers
Near Cell Regional
Telco CO Core DC
tower DC
Edges
Slicing – Edge Stack Edges 5
Network Slicing Phase 2 – Multiple generic NSes Operator software - NFs, Apps,
Mgmt

with application slices and operator provided security

Traditional OSS/BSS hypothesis:
Simpler Slicing Automation
• Allow Enterprises to deploy various kinds
Network Service & Complex App Orchestrator of applications that require ultra low
(NFVO for K8s such as EMCO ) latency, that require low powered
applications etc…

• Appropriate UPFs are configured to steer

the traffic to local application instances via
RU security slices

• Operators also provide security slices (Red)

for the application slices
RU

mMTC

RU eMBB

Device Edge Towers

Near Cell Regional
Telco CO Core DC
tower DC
Edges
Slicing – Edge Stack Edges 6
Dynamic Network Slicing
What is Dynamic Slicing : A way to Create/Delete/Modify Slices on demand basis; Allocate resources from shared pool of
resources; Adjust resources dynamically to meet SLA (Service Assurance)

Questions that come to mind are: When do you need dynamic slices? Which kind of Enterprises go for network slices over
and beyond application slices?

Enterprises/ Organizations Enterprises/ Organizations Enterprises/ Organizations

that have highest security that have highest that don’t trust operator Organizations provide
requirements performance determinism technologies of other managed services to
countries Enterprises
Data leaking concerns with Processing isolation
No Sharing of 5GS NFs and concerns resulting from Bring their own 5GS and
associated services shared NFs associated software

Slicing – Edge Stack

 Operator software - NFs, Apps,

Network Slicing Phase 3 – Dynamic network slices Mgmt

Traditional OSS/BSS
Full fledged Slicing Orchestration
Network Service & Complex App Orchestrator
(NFVO for K8s such as OpenNESS- EMCO)
• Operator with generic slices.
• Operator provides network slices to its
customers (org1, OrgN in this example).

• OrgN provides application slices to its

customers (Enterprise 1 and Enterprise N).
RU Org1 provides app slices to its customers
(In this example, Enterprise X). Operator
Organization N itself has its own customers for application
RU slices (Enterprise Y).

• OrgN gets its own 5GS function too,

Organization 1 whereas Org 1 lets the operator bring the
RU 5GS functions.

Device Edge Towers Generic

Near Cell Regional

Telco CO Core DC
tower DC
Edges
Slicing – Edge Stack Edges 8
 Dynamic slicing framework considerations / Challenges
Considerations Challenges
1. Number of RAN subnet slices can be different from 1. There is no standard way of passing slice information on
number of Core subnet slices. per packet basis.
2. Resource sharing is not just limited to RAN domain & 2. 3GPP introduced ‘NetworkInstance’ as a way to
Core domain. There are resources shared at the communicate the slice information from UPF and CU-UP to
transport (Fronthaul, midhaul and backhaul) and even virtualization layer (or transport layer). But, there is no
at the virtualization layer (K8s and Linux stack). Hence standard way of defining the format of the
resource accounting, allocation and reassignments are networkinstance keeping different underlays in mind.
needed there too for performance isolation and to 3. Virtualization layer (K8s networking) sits in between CNFs
meet SLAs and Transport. There needs to be a way to have
3. Sharing of network functions across subnet slices. interworking layer to bridge the slice information between
Network function shall be able to control resource UP CNFs and transport.
usage among the subnet slices.

E2E Network Slicing shall consider E2E networks including CNFs, Virtualization layer, transport networks

Slicing – Edge Stack

Summary
Some thoughts are given here on the evolution of Network Slicing along with
Application Slices

- No network slicing.
- Application Slices for Edge-computing
- Static network Slices and Edge-computing
- Need for dynamic slicing
- How does dynamic slicing would look like.

Finally, it provided some considerations and challenges.

We will be looking at network slicing implications to OpenNESS-EMCO,

Kubernetes & container runtimes and possibility of extending EMCO to include
Slicing orchestration.
Slicing – Edge Stack 10