2016 IEEE Conference on Standards for Communications and Networking (CSCN)

Network Slice Selection, Assignment and Routing

within 5G Networks
Vinod Kumar Choyi1, Ayman Abdel-Hamid2, Yogendra Shah1, Samir Ferdi1, Alec Brusilovsky1
1
InterDigital Communications Inc., Conshohocken, PA
2
University of Pittsburgh, Pittsburgh, PA
1
{vinod.choyi, yogendra.shah, samir.ferdi, alec.brusilovsky}@interdigital.com, [email protected]

Abstract — 5G Networks are anticipated to provide a diverse set specific service, targeted towards a subset of 5G use cases. A
of services over Network Slices (NS) using Network Function UE’s service level requirements may then be met by one or
Virtualization (NFV) technologies. We propose solutions to more NS. In this evolved architecture model, the desire to
enable NS selection and routing of traffic routing through a NS. match a UE’s service level requirements to appropriate NS(s)
A framework for enabling negotiation, selection and assignment emerges as a new challenging proposal. In this paper, we
of NSs for requesting applications in 5G networks is presented. propose methods to enable a standardized means for NS
A definition for service description has been developed, which is selection and assignment using a service negotiation protocol.
then leveraged for negotiation between an application/user We introduce the notion of a Service Description Document
equipment and the serving network. Virtualized networks or
(SDD) to align a UE’s service requirements with the services
slices of virtualized networks are selected and assigned based on
QCI and security requirements associated with a requested
offered by a SP’s NS.
service. We also describe static and dynamic mechanisms for In addition, due to the distributed nature of cloud
packet routing within NSs. computing, virtualization and SDN technologies, the
underlying network configuration for a NS is expected to be
Index Terms — 5G systems, virtualized networks, security, highly flexible and agile. This brings with it the problem of
NFV, QCI, service provisioning, network slice selection
identifying a best route through a NS to meet the over-arching
requirements for service quality (QoS, security etc). In this
I. INTRODUCTION paper we also propose a protocol for static, dynamic and
The use cases driving the architecture for 5G networks hybrid routing of packets based on an extended service QCI
encompass not only the ongoing need for Enhanced Mobile using Service Function Chaining (SFC) [3] and Network
Broadband (eMBB) connectivity, but also Massive Machine Service Header (NSH) [4].
Type Communications (mMTC) and Ultra-Reliable Critical
Communications (URCC) services [1]. In addition, 5G This paper is organized as follows. Section II provides use
systems will be required to provide transparent support for cases, requirements and prior work pertaining to 5G systems.
legacy User Equipment (UE). Section III gives a complete overview of the service
negotiation framework and flexible routing. Section IV details
While 5G systems will, no doubt, continue to up the ante selected technical characteristics and assessment of the
on the data rate and make uplink data throughput as high as or architecture. Concluding remarks including some notes on
in many use cases exceeding the downlink, we believe that the security are contained in Section VI.
focus will also be on coverage and user experience. The
interest in 5G technologies is already simmering, and the II. BACKGROUND
industry is starting to fund projects looking into such
technologies. With much of the industry viewing the future A. 5G Systems: Use Cases and Business Models
wireless standards as mostly focused on data rates, large
amounts of data, power, processing efficiency, and enabling
new revenue generation (e.g., IoT services), network vendors
and operators are also looking at technologies which will cope
with 1000x traffic growth. The goal of 5G systems
standardization is to not only meet such traffic growth but also
accommodate the diversity of use cases, while taking into
consideration architectural approaches to the deployment of
5G networks that reduce CAPEX and OPEX costs from a
Mobile Network Operators (MNO) or Service Provider (SP)
perspective. Underlying the revolution is the notion that cost
reduction and flexibility for wireless networks may be
achieved, by reducing the dependency on dedicated network
functions residing on dedicated hardware platforms and
switching to generic Commercial-off-the-Shelf (COTS)
platforms and cloud computing utilizing virtualization and Figure 1: Use-cases Supported by 5G
SDN technologies.
Realizing the features of 5G networks requires
The latest efforts towards addressing the needs of 5G consideration along three different dimensions: business
systems, in order to meet the needs for a multiplicity of often models, use-cases and service requirements. Generally
conflicting new services requirements [2], are bringing a new speaking, different MNOs have different business models and
paradigm shift that is rapidly gaining ground. The notion of a different services they wish to render to end users. An MNO
Network Slice (NS) has emerged where a NS provides a can be an asset provider (IaaS, PaaS), connectivity provider

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE MINAS GERAIS. Downloaded on November 29,2023 at 00:05:59 UTC from IEEE Xplore. Restrictions apply.
978-1-5090-3862-6/16/$31.00 ©2016 IEEE
 2016 IEEE Conference on Standards for Communications and Networking (CSCN)
2

(e.g. MVNO), or partner service provider. The different use- Figure 3 depicts NSs for mMTC (IoT) and URCC that
cases dictate new requirements and corresponding features of includes control and user plane functions, created and realized
5G as illustrated in Figure 1 [1]. Finally, the service using network functions, that may be Physical Network
requirements have been expanded, relative to earlier wireless Functions (PNF) or Virtual Network Functions (VNFs) or a
systems, to include reliability, availability, and resilience and combination of both. An MNO may leverage the services of a
emphasis varies with different use cases. Management and Network Orchestration (MANO-like
functionality) to configure SDN controllers which in turn are
B. Network Slicing used for configuration of the switching / routing fabric. An
A NS is defined as a set of features that achieves the Orchestrator is able to instantiate a NS using a static list of
requirements of a certain use-case taking into consideration information elements using a Network Service Descriptor
the capabilities of the MNO(s) that provide the service [5]. It (NSD), which provides a Blueprint to create a NS from a
is envisioned that 5G networks will support vertical use cases collection of various VNFs and their inter-connections [8].
such as intelligent transportation, gaming, IoT, remote
machinery and virtual reality that may have different service The three main uses cases driving 5G networks can be
requirements [1]. In order to be able to support such expanded into a range of QoS Class Identifiers (QCI), which
differentiated applications and services, virtual slices of provide fine grained service selectivity. For example, the
network are intended to cater to the different requirements for eMBB category can be sub-divided into ultra-high speed, high
eMBB, mMTC and URCC as shown in Figure 2. speed, high mobility or low cost where each of these specific
categories may be serviced by different NS specifications.
Thus, with this expansion of the services to be covered in 5G
networks as well as the flexibility of such networks, it is
natural to extend the notion of QCI to enable selection of a
NS(s). The variety of services to be supported by 5G networks
introduces a variety of different requirements, which are to be
handled by NS(s). Such use cases as eMBB, mMTC, and
URCC all require different types of network features in terms
of security and QoS such as latency, resiliency, mobility,
reliability. For example, an eMBB service may have high data
rates, latency may be tolerable with nomadic mobility whereas
in a URCC service, a very low latency, resiliency and security
Figure 2: Requirements for 5G use cases may be critically important. An mMTC communication may
not require features such as handover and mobility
Such virtual slices are a logical instantiation of the management or be demanding in terms of latency. However,
required network resources. A NS may include the Core density of devices may be extremely high.
Network, RAN and even UE resources, i.e. an end-to-end
network service. Using automated management based on NFV C. Related Work
technologies, MNOs and SPs will be able to support such Network service deployments are often coupled to
slicing. NFV technology is a new way to build an end-to-end network topology. For example, use of a firewall requires
network infrastructure with evolving IT virtualization efficient traffic flow through the firewall, which means
technology so as to enable the consolidation of 5G network placing the firewall on the network path (often via creation of
functions onto standard high-volume servers, switches, and VLANs) or architecting the network topology to steer traffic
storage [6]. A complementary technology to the previously through the firewall. Such dependency inhibits the MNO from
mentioned NFV is Software Defined Networking (SDN), optimally utilizing service resources, and reduces flexibility.
which enables realization (e.g. routing and switching) of the SFC utilizes a service-specific overlay that creates the service
control plane, user plane and management plane for 5G. SDN topology. The service overlay provides service function
is a solution for enabling NFV and network programmability, connectivity, built "on top" of the existing network topology.
which are the protocol building blocks for 5G networks [7]. Thus, it allows MNOs to use whatever overlay or underlay
they prefer to create a path between service functions and to
locate service functions in the network as needed. In [9], the
Network Function Decomposition (NFD) notion was
proposed and an algorithm to choose the number of NFDs
such that the network resource utilization is maximized.
Moreover, [10] introduces the concept of a control plane SFC
applied to an OTT webRTC over 4G/LTE.

III. SLICE NEGOTIATION ARCHITECTURE AND

FRAMEWORK
A. Basic Concept
We present a framework and mechanisms to enable an
Application Service Provider (ASP) to provide services based
on a service descriptor (e.g. high mobility or legacy streaming
Figure 3: Slices for mMTC and URCC etc.) associated with an application. The characteristics of the
application are interpreted into detailed information

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE MINAS GERAIS. Downloaded on November 29,2023 at 00:05:59 UTC from IEEE Xplore. Restrictions apply.
 2016 IEEE Conference on Standards for Communications and Networking (CSCN)
3

containing the QCI service requirements to include security Since the diversity of deployments and business case
service requirements of the application. A summary of the scenarios in 5G systems introduces a variability in terms of the
high-level steps that are performed so that data from an level of security required for each of the offered services, we
application on the UE is routed based on the service propose the addition of one or more parameters indicating the
requirement for that applications is provided below: security level for a corresponding service within the SDD.
These security level parameters may be qualitative or
Step 1: NS assignment based on QCI/security requirements quantitative values and may be interpreted as the minimum
using SDD acceptable level of security for specific services. The security
Step 2: Determining SFC and path based on NS / SDD level parameters may then be used to route traffic through a
Step3: Routing through a service function path NS through appropriate VNFs.
B. Service Description Document D. Service Descriptor based Slice Assignment
A Service Description Document (SDD) contains a A detailed slice request, negotiation and assignment
description of the services and their corresponding QCI (e.g., process is defined through the representation of a set of
latency, throughput, security level, etc.) to be provided by a service-specific or application-specific service descriptors
particular NS. QCIs are commonly used to define different incorporated into an SDD.
data bearers that are to be assigned to different sessions. A
service descriptor is introduced to extend the notion of QCI to When a user initiates a session by activating an
support service selection and security. An SDD service application, the application/UE sends an SDD to the network.
descriptor may include additional preferences for a particular This SDD contains the package of services requested by the
NS, e.g., geographical location, hardware used, capacity of the user. The SDD may enable an application or a UE to provide
NS, etc. Table 1 provides an illustrative summary of the QCI information about the application service characteristics and
that may be defined and the corresponding Key Performance provide granular detail about security and other QCI
Indicators (KPIs) as derived from [12]. The first three requirements to a Serving Network (SN). Figure 4 illustrates
parameters in this table (data rate, latency and reliability) are an example SDD, the mechanism by which an application or
inherited from the legacy 3GPP LTE QCI table [1]. However, underlying application-aware service layer or other lower
such parameters are not sufficient to provision the layers is relayed the required service from a UE to a SN.
aforementioned new features and requirements in 5G
networks. For example, as shown in the table, a low Packet
Loss Rate (PLR) value does not guarantee a reliable network.

Figure 4: SDD Containing Detailed Service Information

In this example, the SDD describes the service/security
characteristics of a “MyChatApp” messaging application. The
services associated with the application are legacy
conversational video and voice. The header of the SDD
contains an identifier for the SDD, denoted as “SDD_ID”,
type of application, denoted as Type_of_App, class of
application, denoted as Class_of_App and the required
Table 1: QCI/Security and Associated KPI security level, denoted as Security_Level. The SDD may be
encoded using JSON, XML or using a compressed data
However, by adding the Radio Link Control (RLC) record. The SDD may be protected for integrity and optionally
protocol at the radio link layer (RLC mode – Transparent, for confidentiality. For example, the integrity of the SDD may
Ack. Or UnAck.) and the Hybrid-ARQ retransmissions be provided by creating a digital signature by the issuer using
(retransmissions threshold) at the MAC layer, configured in asymmetric or symmetric cryptographic mechanisms.
different modes can enable more reliable communications. Optionally, the SDD may be protected for confidentiality and
Security attributes relating to integrity, confidentiality, and privacy by encrypting the SDD and enveloping the SDD into
user privacy are added to the QCI and the associated KPI a secure object (e.g. JWE object). An SDD therefore may
values. Other security attributes relating to authorization contain additional security specific data e.g., the algorithm
procedures, authentication mechanisms protocols and used for creating the signature, denoted as “Algorithm”, the
algorithm complexity (algorithm strength, credential strength) “Issuer” of the SDD, a credential identifier, denoted as
may also be taken into consideration. For the sake of brevity, “Cred_ID” and a “Signature”, which may be a digital
we have only shown a few of the attributes in the table. signature or an authentication tag depending upon the type of
C. Security as a Parameter for the SDD algorithm that is being used. In the example illustrated above,

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE MINAS GERAIS. Downloaded on November 29,2023 at 00:05:59 UTC from IEEE Xplore. Restrictions apply.
 2016 IEEE Conference on Standards for Communications and Networking (CSCN)
4

a digital signature on the SDD is generated using the private More clearly, the slice selection flow-chart can be
key of the “Issuer” (e.g. My Communications Corp). For the articulated in the following steps:
sake of simplicity and brevity we only illustrate some of the
possible data associated with an SDD. Other information 1. A request message containing an SDD is received by the
elements that may be included as part of the SDD may support SN from an UE. The Request may contain a Requested
fine tuning of the NS selection process or steer a UE towards SDD (SDDR). The SDDR may contain one or more SDDs.
a preferred operator NS. (e.g., an array of SDDR) The SN makes a determination
whether a lower-layer connection Request is made by a
legacy UE (e.g., 4G device) or not. If the UE was
determined to be a legacy UE, then the process jump to
Step 9 or in some cases based upon policy, the process
may be terminated. This step may be useful in order to
mitigate masquerading attacks (e.g. UE pretending to be
a 5G device).
2. The SN makes a determination whether the UE is
authorized to perform such a Request by retrieving the
UE/Subscriber Profile (SProf), which contains
information about the UE/Subscriber which is stored in a
database within the SN or in a trusted third-party network.
If the UE had not been authorized then the flow jumps to
step 15.
3. If the UE has been deemed to be authorized to perform
the Request, then the SN checks to see if the UE had
provided a Requested SDD (SDDR). If UE did not
provide SDDR, then the SN performs Step 10.
4. If an SDDR was present in the Request message, then the
SN checks to see if it possess an Offered SDD (SDDO),
that fulfils the SDDR. It must be noted that an SDDR or
SDDO may be made up of one or more SDDs.
5. A check is performed to see if there is an exact match
between the SDDR and the SDDO. An example of a perfect
match would be if a UE had requested an SDDR for
“MyChatApp” as illustrated in Figure 4 and the SN has a
NS (SDDo) specifically tailored for “MyChatApp”. Note
that multiple NSs may be required to fulfill the SDDr for
“MyChatApp,” each service may be allocated to a
Figure 5: Flow-chart for Slice Negotiation
different operator of a NS. The specification of the SDDR
In 5G networks, it is anticipated that different services may would provide the SN with a means to select at an
be offered by different SPs. In addition, the infrastructure may individual NS level towards an exact match.
be owned by one party and leased to an MNO under a 6. If there was an exact match between the SDDR and the
network-as-a-service business model through for example, a SDDO, then additional authorization checks (e.g. risk-
contractual agreement. Standardized or pre-provisioned based authentication) such as those described in [11] may
network services may be requested by a UE by including an be carried out.
SDD_ID within its request message to the SN. An alias may 7. The SN determines if the UE/subscriber authentication /
be included instead of an SDD_ID (e.g., “MyChatApp”) in the authorization is successful. If not then go to step 11.
request message sent by the UE to the SN. Note that such an 8. If the authorization checks have been successfully
SDD alias translation into service requirements can work if
completed, then the SN checks to see if the UE had
such SDD_ID or SDD alias are standardized. As illustrated in
requested for catalog of services offered by the SN.
the 2nd column of Table 1, additional examples of service
descriptors are “legacy conversational video” and “legacy 9. If the UE did not Request for an entire catalog of services
conversational voice” or default services (e.g., legacy 4G offered, then a NS that matches the SDDR is assigned to
service, Emergency Calling service, credential provisioning the UE/application.
service, etc.), which may be available for selection in both, 10. A process for obtaining the SDDR from the SProf is
home and roaming scenarios. performed and results are provided in order to determine
the default SDD that can be assigned to the UE.
When a network receives an SDD, it attempts to match the 11. If an SDDo does not match the SDDR as determined in
SDD characteristics with the network’s slices. The NS step 4 or if the UE has not been authorized based on a
selection process can be summarized by means of a flowchart determination at step 7, then a service request denial is
as illustrated in Figure 5. The mechanisms described and
sent by the SN to the UE.
illustrated involve a UE and a SN performing the slice
12. In case an UE has not been authorized to perform Request
negotiations.
for a NS as determined in step 2, then the UE may seek

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE MINAS GERAIS. Downloaded on November 29,2023 at 00:05:59 UTC from IEEE Xplore. Restrictions apply.
 2016 IEEE Conference on Standards for Communications and Networking (CSCN)
5

other mechanisms to obtain authorization (e.g. using bound to a fixed topology and does not adapt well to elastic
payment/subscription option). service environments enabled by virtualization. Hence, a
13. If the SDDO does not match the SDDR, then UE may topology-independent approach can enable and exploit the
determine if the SDDO is acceptable or not. dynamic nature of deploying the classifiers more flexibly
14. If the SDDO was acceptable to the UE, then the UE may across a network. Figure 6 illustrates the static and dynamic
determine the right SDDR from the SDDO. In some cases, models. A hybrid model is a combination of static and
the SDDR may be the complete SDDO. dynamic models.
15. If the SDDO is deemed not to be acceptable to the UE,
then the UE may make a determination to request for a
catalog of services to the SN.
16. If a determination is made that a catalog has to be
requested then the UE forms a Request message for
obtaining the catalog of services from the SN.
17. If the SN determines that a catalog has been requested by
the UE in step 8, then the SN offers its catalog of services.
18. Based on the catalog(s) services offered by the SN, the
UE selects SDDR from it.
19. If an SDDR has been selected by the UE, a request sent by
the UE containing the SDDR is received by the SN. The
SDDR may have been selected using the SDDo in step 14
or using the catalog at step 14.

A UE can negotiate for a NS either based on its current

SDD request only or the network can take into consideration Figure 6: Static and Dynamic Paths
the UE/Subscriber’s profile jointly with the current SDD
request sent by the UE. Alternatively, an ASP can perform the The determination of static and dynamic SFPs are
negotiations on behalf of the UE with the SN. In a back-end explained in the following sections.
NS negotiation scenario such as the latter example, (i.e. ASP
A. Static SFP
requests SN for a NS using an SDD), we envision the
possibility - in the absence of a particular NS with required In this deployment model, the full end-to-end network path
characteristics or capacity, in the case of a high demand for is predetermined at the ingress classifier and cannot be
such a NS - for the SN to offer the possibility to create a NS. modified by intermediate nodes/entities. As depicted in Figure
Such a NS may be created by way of translation or mapping 6(a), a static route is predetermined by the classifier and
of the SDD QCI and security requirements into an appropriate cannot be altered afterwards even if one of the VNFs is
format such as NSD. This Blueprint can then be interpreted by congested, a VNF or a link failure is identified, or if a
an Orchestration system to instantiate the NS. compromised VNF has been detected. A packet that is
classified using a static path would traverse the following
IV. DETERMINATION OF ROUTING PATHS sequence of functions: classifier A ĺ VNF1 ĺ VNF2a ĺ
VNF3a ĺ VNF4a ĺ VNF5 ĺ VNFn. In this static model,
SFCs provide flexibility in provisioning UE service network service deployments are coupled with network
requirements by chaining the required service functions (SFs) topology. Such a statically established dependency imposes
independently of the network topology. The network constraints on service delivery, potentially inhibits the MNO
management function translates an SDD from a from optimally utilizing service resources and reduces service
UE/application and assigns an appropriate NS. Based on the flexibility. The configuration of the paths on to the routing
NS information, the network determines the SFC. There may infrastructure is performed using the services of SDN
be one or more SFCs that may satisfy a NS requirement. For controllers. Packets are routed using routing
example, an SFC can be a chain of user plane functions information/intelligence within the network (e.g.,
comprising many VNFs. The SFC may also be separately configuration of the SDN controllers) or routing information
applied to control plane functions such as handover or incorporated into the packets. We briefly describe the
authentication signaling, mobility management, etc. We approaches.
propose to consider the QCI as well as the security level as
input parameters when determining the exact VNF instances Routing by means of Pre-configured Static SFP
to be included in a specific SFC. A classifier located at the The classifier reads the packet header (e.g. Slice-ID
ingress (e.g. an ingress router/switch/server) of the network information within the packet header) to determine the full
classifies the SFC to pre-determine the full service function SFP. The SFC is classified and mapped to an SFP and the
path (SFP) of the flow. The mapping of a logical SFC to a network path determined and set up based on intelligence in
physical SFP can be realized by using either a static SFC the network (e.g., in the SDN controllers.) The network path
mapping, a dynamic SFC mapping, or a hybrid mapping. A is not, typically, reclassified.
static SFC mapping approach is topology-dependent such that
an ingress classifier is placed at the network entrance to Routing using Static Path Defined within Packet
classify the SFC accompanied with the received packet and
In this approach, the packet or datagram emanating from
define a full consistent physical path. However, this type of
the application in the UE plays a vital role in its own routing.
service function deployment model is relatively static, and

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE MINAS GERAIS. Downloaded on November 29,2023 at 00:05:59 UTC from IEEE Xplore. Restrictions apply.
 2016 IEEE Conference on Standards for Communications and Networking (CSCN)
6

assumed that the CF has been pre-configured with

information about NSs, SFCs associated within each NS
and optionally connection mechanism (e.g. protocol)
from the CF to the next hop (e.g. SDN switch). As part of
this process, the traffic from the UE is transported to the
CF that is part of the SN.
2. The CF may use the Slice-Id in order to perform a look-
up of the list of SFCs that make up the SFP. The CF
fetches information about the NS and (e.g., MNO)
policies associated with the NS.
3. Based upon the policies and information obtained by the
CF then encapsulates an NSH to the packet header. The
CF then adds the list of SFCs or SFP components within
the NSH. It must be noted that the SFP components may
be addressable as: IP address and port number, URL and
port number, VNF-Id (which may be translated to
addressable endpoints), etc. The CF identifies the next
VNF that must be targeted for this packet.
4. The CF forwards the packet to SDN Switch 1. The
determination of the forwarding of the packet is based
Figure 7: Packet Routing using Static SFPs upon the routing information that was populated by the
SDN controller during Step 0.
In order to enable flexibility and dynamic function
selection for the SFC deployment, some information is added 5. Based on the information that has been provided within
to the packet’s header to indicate the SFP. The classifier reads the NSH, the SDN Switch 1 determines the VNF within
the packet header information to determine the full SFP and the SFC chain and determines the appropriate SDN
constructs header information for the packet. This Switch that could lead the packet to the appropriate VNF.
functionality can be included into some of the currently used 6. The Switch 1 forwards the packet/datagram to SDN
L2/L3 network protocols, e.g., source-routing headers. In Switch 2.
order to achieve this QCI based routing, we propose to use an 7. Switch 2 inspects the NSH and determines that the next
enhancement of the NSH, which is being standardized within VNF that must service the packet can be accessed via it.
IETF [4]. In this proposal the NSH header contains service The switch translates the SPI and the SI into the next-hop,
path information and the QCI which can be included within i.e., Next NF to be executed
the optional metadata information. The NSH is then added to 8. Switch 2 forwards the packet to the VNF1. The
a packet or frame. The NSH contains service path mechanism by which the packet is transmitted to the
identification information that can be used by a SFF to forward VNF1 may involve higher-layer protocols (e.g. HTTP).
the packets through appropriate VNFs. As the packet traverses 9. The VNF1 processes the packet. Processing the packet
through the network, each VNF can read the NSH header may involve performing any required security, session
information in order to determine the next hop. i.e., the next management operations, etc., based on the data contained
VNF that has to be executed. An example call flow that within the packet and based on the type of network
describes routing of the packet by means of a static path and function being performed by the VNF1.
using NSH is illustrated in Figure 7.
10. Once the packet has been processed, the VNF1 forwards
The steps involved in routing are detailed below: an acknowledgement of the received packet to Switch 2
or to any other routing/switch (e.g., a default Switch),
0. Information about NSs and/or SFC are populated by the
which then forwards to the appropriate next VNF target
SFC Routing Information Dissemination Function
function in the SFC.
(SRIDF) to the Classifier Function (CF) and the SDN
11. Switch 2 modifies the NSH and determines the next hop
Controllers. The information disseminated to the CF
VNF and process is repeated till the SFP is traversed.
would include the details on the SFC for each NS. It may
also include policies associated with the creating and B. Dynamic SFP
usage of NSH, which may be based upon the SFC for that In this approach, the SFC can be classified, path reclassified
NS. The information that is provided to the SDN at some instances, along the network path according to the
Controller may be very detailed and granular information network status, i.e., VNF congestion, a link failure etc. Since
(e.g. policies on routing of packets/datagram based on the static topology information is increasingly less viable due to
NSH, default routing etc...). The information would scaling, tenancy and complexity reasons, a dynamic SFC
provide details on how the VNFs may be reached and approach becomes attractive for 5G systems. As illustrated in
default routing information in case a VNF cannot Figure 6(b), classifiers (A, VNF3a) are placed throughout the
reached. network to remap some of the SFCs in case any of the SFPs
1. A packet from a UE, identified by a UE-Id and optionally are no longer consistent with the QCI of the provisioned NS.
a Slice-Id arrives at a CF. It is assumed that the initial For example, if VNF2b and VNF4a are degraded (e.g. due to
attachment point with a SN is provided to the UE and the congestion), then the classifier A and VNF3a that have a
SN forwards the packet from the UE to the CF. It is also reclassification capability, can re-route packets using a newer

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE MINAS GERAIS. Downloaded on November 29,2023 at 00:05:59 UTC from IEEE Xplore. Restrictions apply.
 2016 IEEE Conference on Standards for Communications and Networking (CSCN)
7

route. A packet is then routed using the newer path that through NFV and SDN, is selected and assigned based on the
traverses the following sequence of functions: classifier A ĺ QCI/security requirements of the service descriptor.
VNF1 ĺ VNF2b ĺ VNF3a ĺ VNF4b ĺ VNF5 ĺVNFn. Subsequently, static or dynamic routing mechanisms are used
The packet may carry routing information/intelligence to treat data packets according to the QCI/security values and
relating to the SFP. The network dynamically determines the to flexibly select network functions and service function paths
actual SFP to be navigated based upon a QCI indication or through a NS.
label included within the packet/datagram. A Classifier is not
present, unlike the static path case, and, therefore, a network REFERENCES
path is not pre-determined. At each hop, a next hop VNF is
[1] Alliance, N. G. M. N. "5G white paper." Next Generation
determined provided that the VNF that is selected is part of the
SFC and also meets or exceeds the QCI requirements. Mobile Networks, White paper (2015).
[2] https://www.etsi.org/deliver/etsi_ts/132400_132499/132
Network conditions are susceptible to change over time 425/09.05.00_60/ts_132425v090500p.pdf
due to congestion, failures, attacks or other vulnerabilities and, [3] Service Function Chaining (SFC) Architecture, IETF
therefore, flows may have to be re-directed to alternative RFC 7665, October 2015
routes. In this paper, we propose the notion of hop-by-hop [4] Network Service Header, IETF draft-ietf-sfc-nsh-04.txt,
QCI. Instead of pre-selecting the full path that meets the QCI https://datatracker.ietf.org/doc/draft-ietf-sfc-nsh/, March
requirements for the network flow as determined by a static 2016
SFP, the notion of hop-by-hop selection enables the network [5] “Description of Network Slicing Concept,”
flow to be classified in a hop-by-hop manner (i.e., at each
https://www.ngmn.org/uploads/media/160113_Network
switch/entity) to route packets through the best next hop
_Slicing_v1_0.pdf
according to the QCI requirements of the flow (including
security level requirements). This can be attained by placing a [6] NFVISG ETSI, “Network functions virtualization, white
number of SFC classifiers throughout the network. paper,” 2012.
[7] “Open Networking Foundation. (2012, Apr.). software-
In the proposed hop-by-hop QCI solution, we assume defined networking: The new norm for
SDN-based networks at both the RAN and the core networks. networks[online],”
An SFC classifier switch is deployed at the ingress point of the https://www.opennetworking.org/images/stories/downlo
RAN network. This classifier is in charge of allocating the full ads/sdn-resources/white-papers/wp-sdn-newnorm.pdf
network path that corresponds to the SFC. It is worth noting [8] ETSI GS NFV-MAN 001; Management and
that a classifier has the privilege to update the SFC, unlike a
Orchestration
Service Function Forwarder (SFF) that can only forward the
[9] A. Banchs et al., A novel radio multiservice adaptive
packets based on the destination that corresponds to the
executing VNF that was predetermined by a classifier. network architecture for 5g networks. In Vehicular
Technology Conference (VTC Spring), 2015 IEEE 81st,
It may be possible that a system may use dynamic SFP pages 1–5. IEEE, 2015.
routing for the initial session setup or for configuration of a [10] Samuel Jero et al., Dynamic control of real-time
NS, which is then used to create a static SFP. Very first communications (RTC) using SDN: A case study of a 5G
packets for a particular session (for example during session end-to-end service, 2016 IEEE/IFIP Network Operations
setup) may be routed following a hop by hop classification and and Management Symposium (NOMS), April 2016, pp
switching mechanism using NSH insertion and lookup. This 895-900.
initial traffic setup would then lead to the establishment of an [11] Yogendra Shah, Vinod Choyi, Andreas Schmidt,
SFP, after which subsequent packets could be routed using
Lakshmi Subramanian, Multi-Factor Authentication as a
regular network based routing i.e. as soon as an SFP is traced
Service (MFAaaS), 2015 3rd IEEE International
once, normal SDN flow control can follow without the need
for the NSH. Conference on Mobile Cloud Computing, Services and
Engineering.
C. Hybrid Path Routing [12] 3GPP TS 23.203, Policy and Control Charging
In an alternative hybrid network approach, a part of the Architecture
SFP is established using a static path, whereas anther part of
the SFP is realized using dynamic path mechanisms. This is
particularly important in cases where a CP/UP/MP associated
with a critical service may require the use of a specific set of
VNFs, while the non-critical service may be facilitated by
other VNFs using dynamic routing.

V. CONCLUSION
We have presented a framework for providing customized
virtualized networks in 5G based on a service descriptor
document, which includes QCI and security requirements. The
framework enables negotiation, selection, and assignment of
virtualized networks to requesting applications in 5G systems.
A definition for a service descriptor which is used to leverage
negotiation between an application/UE and the serving
network has been presented. A virtualized NS, realized

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE MINAS GERAIS. Downloaded on November 29,2023 at 00:05:59 UTC from IEEE Xplore. Restrictions apply.