RFC 8568

Internet Research Task Force (IRTF)                        CJ. Bernardos
Request for Comments: 8568                                          UC3M
Category: Informational                                        A. Rahman
ISSN: 2070-1721                                             InterDigital
                                                              JC. Zuniga
                                                           LM. Contreras
                                                               P. Aranda
                                                                P. Lynch
                                                   Keysight Technologies
                                                              April 2019

               Network Virtualization Research Challenges


   This document describes open research challenges for network
   virtualization.  Network virtualization is following a similar path
   as previously taken by cloud computing.  Specifically, cloud
   computing popularized migration of computing functions (e.g.,
   applications) and storage from local, dedicated, physical resources
   to remote virtual functions accessible through the Internet.  In a
   similar manner, network virtualization is encouraging migration of
   networking functions from dedicated physical hardware nodes to a
   virtualized pool of resources.  However, network virtualization can
   be considered to be a more complex problem than cloud computing as it
   not only involves virtualization of computing and storage functions
   but also involves abstraction of the network itself.  This document
   describes current research and engineering challenges in network
   virtualization including the guarantee of quality of service,
   performance improvement, support for multiple domains, network
   slicing, service composition, device virtualization, privacy and
   security, separation of control concerns, network function placement,
   and testing.  In addition, some proposals are made for new activities
   in the IETF and IRTF that could address some of these challenges.
   This document is a product of the Network Function Virtualization
   Research Group (NFVRG).

Bernardos, et al.             Informational                     [Page 1]

RFC 8568       Network Virtualization Research Challenges     April 2019

Status of This Memo

   This document is not an Internet Standards Track specification; it is
   published for informational purposes.

   This document is a product of the Internet Research Task Force
   (IRTF).  The IRTF publishes the results of Internet-related research
   and development activities.  These results might not be suitable for
   deployment.  This RFC represents the consensus of the Network
   Function Virtualization Research Group of the Internet Research Task
   Force (IRTF).  Documents approved for publication by the IRSG are not
   candidates for any level of Internet Standard; see Section 2 of RFC

   Information about the current status of this document, any errata,
   and how to provide feedback on it may be obtained at

Copyright Notice

   Copyright (c) 2019 IETF Trust and the persons identified as the
   document authors.  All rights reserved.

   This document is subject to BCP 78 and the IETF Trust's Legal
   Provisions Relating to IETF Documents
   (https://trustee.ietf.org/license-info) in effect on the date of
   publication of this document.  Please review these documents
   carefully, as they describe your rights and restrictions with respect
   to this document.

Bernardos, et al.             Informational                     [Page 2]

RFC 8568       Network Virtualization Research Challenges     April 2019

Table of Contents

   1.  Introduction and Scope  . . . . . . . . . . . . . . . . . . .   4
   2.  Terminology . . . . . . . . . . . . . . . . . . . . . . . . .   4
   3.  Background  . . . . . . . . . . . . . . . . . . . . . . . . .   6
     3.1.  Network Function Virtualization . . . . . . . . . . . . .   6
     3.2.  Software-Defined Networking . . . . . . . . . . . . . . .   9
     3.3.  ITU-T Functional Architecture of SDN  . . . . . . . . . .  13
     3.4.  Multi-Access Edge Computing . . . . . . . . . . . . . . .  15
     3.5.  IEEE 802.1CF (OmniRAN)  . . . . . . . . . . . . . . . . .  15
     3.6.  Distributed Management Task Force (DMTF)  . . . . . . . .  15
     3.7.  Open-Source Initiatives . . . . . . . . . . . . . . . . .  16
   4.  Network Virtualization Challenges . . . . . . . . . . . . . .  18
     4.1.  Overview  . . . . . . . . . . . . . . . . . . . . . . . .  18
     4.2.  Guaranteeing Quality of Service . . . . . . . . . . . . .  18
       4.2.1.  Virtualization Technologies . . . . . . . . . . . . .  18
       4.2.2.  Metrics for NFV Characterization  . . . . . . . . . .  19
       4.2.3.  Predictive Analysis . . . . . . . . . . . . . . . . .  20
       4.2.4.  Portability . . . . . . . . . . . . . . . . . . . . .  20
     4.3.  Performance Improvement . . . . . . . . . . . . . . . . .  21
       4.3.1.  Energy Efficiency . . . . . . . . . . . . . . . . . .  21
       4.3.2.  Improved Link Usage . . . . . . . . . . . . . . . . .  21
     4.4.  Multiple Domains  . . . . . . . . . . . . . . . . . . . .  22
     4.5.  5G and Network Slicing  . . . . . . . . . . . . . . . . .  22
       4.5.1.  Virtual Network Operators . . . . . . . . . . . . . .  23
       4.5.2.  Extending Virtual Networks and Systems to the
               Internet of Things  . . . . . . . . . . . . . . . . .  24
     4.6.  Service Composition . . . . . . . . . . . . . . . . . . .  25
     4.7.  Device Virtualization for End Users . . . . . . . . . . .  27
     4.8.  Security and Privacy  . . . . . . . . . . . . . . . . . .  27
     4.9.  Separation of Control Concerns  . . . . . . . . . . . . .  29
     4.10. Network Function Placement  . . . . . . . . . . . . . . .  29
     4.11. Testing . . . . . . . . . . . . . . . . . . . . . . . . .  30
       4.11.1.  Changes in Methodology . . . . . . . . . . . . . . .  30
       4.11.2.  New Functionality  . . . . . . . . . . . . . . . . .  31
       4.11.3.  Opportunities  . . . . . . . . . . . . . . . . . . .  32
   5.  Technology Gaps and Potential IETF Efforts  . . . . . . . . .  33
   6.  NFVRG Focus Areas . . . . . . . . . . . . . . . . . . . . . .  34
   7.  IANA Considerations . . . . . . . . . . . . . . . . . . . . .  35
   8.  Security Considerations . . . . . . . . . . . . . . . . . . .  35
   9.  Informative References  . . . . . . . . . . . . . . . . . . .  35
   Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . .  41
   Authors' Addresses  . . . . . . . . . . . . . . . . . . . . . . .  41

Bernardos, et al.             Informational                     [Page 3]

RFC 8568       Network Virtualization Research Challenges     April 2019

1.  Introduction and Scope

   The telecommunications sector is experiencing a major revolution that
   will shape the way networks and services are designed and deployed
   for the next few decades.  In order to cope with continuously
   increasing demand and cost, network operators are taking lessons from
   the IT paradigm of cloud computing.  This new approach of
   virtualizing network functions will enable multi-fold advantages by
   moving communication services from bespoke hardware in the operator's
   core network to Commercial Off-The-Shelf (COTS) equipment distributed
   across data centers.

   Some of the network virtualization mechanisms that are being
   considered include the following: sharing of network infrastructure
   to reduce costs, virtualization of core and edge servers/services
   running in data centers as a way of supporting their load-aware
   elastic dimensioning, and dynamic energy policies to reduce the
   electricity consumption.

   This document presents research and engineering challenges in network
   virtualization that need to be addressed in order to achieve these
   goals, spanning from pure research and engineering/standards space.
   The objective of this memo is to document the technical challenges
   and corresponding current approaches and to expose requirements that
   should be addressed by future research and standards work.

   This document represents the consensus of the Network Function
   Virtualization Research Group (NFVRG).  It has been reviewed by the
   RG members active in the specific areas of work covered by the

2.  Terminology

   The following terms used in this document are defined by the ETSI
   Network Function Virtualization (NFV) Industrial Study Group (ISG)
   [etsi_gs_nfv_003], the Open Networking Foundation (ONF) [onf_tr_521],
   and the IETF [RFC7426] [RFC7665]:

   Application Plane:  The collection of applications and services that
      program network behavior.

   Control Plane (CP):  The collection of functions responsible for
      controlling one or more network devices.  The CP instructs network
      devices with respect to how to process and forward packets.  The
      control plane interacts primarily with the forwarding plane and,
      to a lesser extent, with the operational plane.

Bernardos, et al.             Informational                     [Page 4]

RFC 8568       Network Virtualization Research Challenges     April 2019

   Forwarding Plane (FP):  The collection of resources across all
      network devices responsible for forwarding traffic.

   Management Plane (MP):  The collection of functions responsible for
      monitoring, configuring, and maintaining one or more network
      devices or parts of network devices.  The management plane is
      mostly related to the operational plane (it is related less to the
      forwarding plane).

   NFV Infrastructure (NFVI):  Totality of all hardware and software
      components that build up the environment in which VNFs are

   NFV Management and Orchestration (NFV-MANO):  Functions collectively
      provided by NFVO, VNFM, and VIM.

   NFV Orchestrator (NFVO):  Functional block that manages the Network
      Service (NS) life cycle and coordinates the management of NS life
      cycle, VNF life cycle (supported by the VNFM) and NFVI resources
      (supported by the VIM) to ensure an optimized allocation of the
      necessary resources and connectivity.

   Operational Plane (OP):  The collection of resources responsible for
      managing the overall operation of individual network devices.

   Physical Network Function (PNF):  Physical implementation of a
      network function in a monolithic realization.

   Service Function Chain (SFC):  For a given service, the abstracted
      view of the required service functions and the order in which they
      are to be applied.  This is somehow equivalent to the Network
      Function Forwarding Graph (NF-FG) at ETSI.

   Service Function Path (SFP):  The selection of specific service
      function instances on specific network nodes to form a service
      graph through which an SFC is instantiated.

   Virtualized Infrastructure Manager (VIM):  Functional block that is
      responsible for controlling and managing the NFVI compute,
      storage, and network resources, usually within one infrastructure
      operator's domain.

   Virtualized Network Function (VNF):  Implementation of a Network
      Function that can be deployed on a Network Function Virtualization
      Infrastructure (NFVI).

   Virtualized Network Function Manager (VNFM):  Functional block that
      is responsible for the life-cycle management of VNF.

Bernardos, et al.             Informational                     [Page 5]

RFC 8568       Network Virtualization Research Challenges     April 2019

3.  Background

   This section briefly describes some basic background technologies as
   well as other Standards Developing Organizations (SDOs) and open-
   source initiatives working on network virtualization or related

3.1.  Network Function Virtualization

   The ETSI ISG Network Function Virtualization (NFV) is a working group
   that, since 2012, has aimed to evolve quasi-standard IT
   virtualization technology to consolidate many network equipment types
   into industry standard high-volume servers, switches, and storage.
   It enables implementing network functions in software that can run on
   a range of industry-standard server hardware and can be moved to, or
   loaded in, various locations in the network as required, without the
   need to install new equipment.  The ETSI NFV is one of the
   predominant NFV reference framework and architectural footprints
   [nfv_sota_research_challenges].  The ETSI NFV framework architecture
   is composed of three domains (Figure 1):

   o  Virtualized Network Function, running over the NFVI.

   o  NFVI, including the diversity of physical resources and how these
      can be virtualized.  NFVI supports the execution of the VNFs.

   o  NFV Management and Orchestration, which covers the orchestration
      and life-cycle management of physical and/or software resources
      that support the infrastructure virtualization, and the life-cycle
      management of VNFs.  NFV Management and Orchestration focuses on
      all virtualization-specific management tasks necessary in the NFV

Bernardos, et al.             Informational                     [Page 6]

RFC 8568       Network Virtualization Research Challenges     April 2019

   +-------------------------------------------+  +---------------+
   |   Virtualized Network Functions (VNFs)    |  |               |
   |  -------   -------   -------   -------    |  |               |
   |  |     |   |     |   |     |   |     |    |  |               |
   |  | VNF |   | VNF |   | VNF |   | VNF |    |  |               |
   |  |     |   |     |   |     |   |     |    |  |               |
   |  -------   -------   -------   -------    |  |               |
   +-------------------------------------------+  |               |
                                                  |               |
   +-------------------------------------------+  |               |
   |         NFV Infrastructure (NFVI)         |  |      NFV      |
   | -----------    -----------    ----------- |  |  Management   |
   | | Virtual |    | Virtual |    | Virtual | |  |      and      |
   | | Compute |    | Storage |    | Network | |  | Orchestration |
   | -----------    -----------    ----------- |  |               |
   | +---------------------------------------+ |  |               |
   | |         Virtualization Layer          | |  |               |
   | +---------------------------------------+ |  |               |
   | +---------------------------------------+ |  |               |
   | | -----------  -----------  ----------- | |  |               |
   | | | Compute |  | Storage |  | Network | | |  |               |
   | | -----------  -----------  ----------- | |  |               |
   | |          Hardware resources           | |  |               |
   | +---------------------------------------+ |  |               |
   +-------------------------------------------+  +---------------+

                       Figure 1: ETSI NFV Framework

   The NFV architectural framework identifies functional blocks and the
   main reference points between such blocks.  Some of these are already
   present in current deployments, whilst others might be necessary
   additions in order to support the virtualization process and
   consequent operation.  The functional blocks are (Figure 2):

   o  Virtualized Network Function (VNF)

   o  Element Management (EM)

   o  NFV Infrastructure, including: Hardware and virtualized resources
      as well as the Virtualization Layer.

   o  Virtualized Infrastructure Manager(s) (VIM)

   o  NFV Orchestrator

   o  VNF Manager(s)

   o  Service, VNF and Infrastructure Description

Bernardos, et al.             Informational                     [Page 7]

RFC 8568       Network Virtualization Research Challenges     April 2019

   o  Operational Support Systems and Business Support Systems (OSS and

   +-------------------------------------------+  | ----------------   |
   |                 OSS/BSS                   |  | | NFV          |   |
   +-------------------------------------------+  | | Orchestrator +-- |
                                                  | ---+------------ | |
   +-------------------------------------------+  |    |             | |
   |  ---------     ---------     ---------    |  |    |             | |
   |  | EM 1  |     | EM 2  |     | EM 3  |    |  |    |             | |
   |  ----+----     ----+----     ----+----    |  | ---+----------   | |
   |      |             |             |        |--|-|    VNF     |   | |
   |  ----+----     ----+----     ----+----    |  | | manager(s) |   | |
   |  | VNF 1 |     | VNF 2 |     | VNF 3 |    |  | ---+----------   | |
   |  ----+----     ----+----     ----+----    |  |    |             | |
   +------|-------------|-------------|--------+  |    |             | |
          |             |             |           |    |             | |
   +------+-------------+-------------+--------+  |    |             | |
   |         NFV Infrastructure (NFVI)         |  |    |             | |
   | -----------    -----------    ----------- |  |    |             | |
   | | Virtual |    | Virtual |    | Virtual | |  |    |             | |
   | | Compute |    | Storage |    | Network | |  |    |             | |
   | -----------    -----------    ----------- |  | ---+------       | |
   | +---------------------------------------+ |  | |        |       | |
   | |         Virtualization Layer          | |--|-| VIM(s) +-------- |
   | +---------------------------------------+ |  | |        |         |
   | +---------------------------------------+ |  | ----------         |
   | | -----------  -----------  ----------- | |  |                    |
   | | | Compute |  | Storage |  | Network | | |  |                    |
   | | | hardware|  | hardware|  | hardware| | |  |                    |
   | | -----------  -----------  ----------- | |  |                    |
   | |          Hardware resources           | |  |  NFV Management    |
   | +---------------------------------------+ |  | and Orchestration  |
   +-------------------------------------------+  +--------------------+

                 Figure 2: ETSI NFV Reference Architecture

Bernardos, et al.             Informational                     [Page 8]

RFC 8568       Network Virtualization Research Challenges     April 2019

3.2.  Software-Defined Networking

   The Software-Defined Networking (SDN) paradigm pushes the
   intelligence currently residing in the network elements to a central
   controller implementing the network functionality through software.
   In contrast to traditional approaches, in which the network's control
   plane is distributed throughout all network devices, with SDN, the
   control plane is logically centralized.  In this way, the deployment
   of new characteristics in the network no longer requires complex and
   costly changes in equipment or firmware updates, but only a change in
   the software running in the controller.  The main advantage of this
   approach is the flexibility it provides operators to manage their
   network, i.e., an operator can easily change its policies on how
   traffic is distributed throughout the network.

   One of the most well-known protocols for the SDN control plane
   between the central controller and the networking elements is the
   OpenFlow Protocol (OFP), which is maintained and extended by the Open
   Network Foundation (ONF) <https://www.opennetworking.org/>.
   Originally, this protocol was developed specifically for IEEE 802.1
   switches conforming to the ONF OpenFlow Switch specification
   [OpenFlow].  As the benefits of the SDN paradigm have reached a wider
   audience, its application has been extended to more complex scenarios
   such as wireless and mobile networks.  Within this area of work, the
   ONF is actively developing new OFP extensions addressing three key
   scenarios: (i) wireless backhaul, (ii) cellular Evolved Packet Core
   (EPC), and (iii) unified access and management across enterprise
   wireless and fixed networks.

Bernardos, et al.             Informational                     [Page 9]

RFC 8568       Network Virtualization Research Challenges     April 2019

   | -------  |
   | |Oper.|  |            O
   | |Mgmt.|  |<........> -+- Network Operator
   | |Iface|  |            ^
   | -------  |      +----------------------------------------+
   |          |      | +------------------------------------+ |
   |          |      | | ---------  ---------     --------- | |
   |--------- |      | | | App 1 |  | App 2 | ... | App n | | |
   ||Plugins| |<....>| | ---------  ---------     --------- | |
   |--------- |      | | Plugins                            | |
   |          |      | +------------------------------------+ |
   |          |      | Application Plane                      |
   |          |      +----------------------------------------+
   |          |                         A
   |          |                         |
   |          |                         V
   |          |      +----------------------------------------+
   |          |      | +------------------------------------+ |
   |--------- |      | |     ------------  ------------     | |
   || Netw. | |      | |     | Module 1 |  | Module 2 |     | |
   ||Engine | |<....>| |     ------------  ------------     | |
   |--------- |      | | Network Engine                     | |
   |          |      | +------------------------------------+ |
   |          |      | Control Plane                          |
   |          |      +----------------------------------------+
   |          |                         A
   |          |                         |
   |          |                         V
   |          |      +----------------------------------------+
   |          |      |  +--------------+   +--------------+   |
   |          |      |  | ------------ |   | ------------ |   |
   |----------|      |  | | OpenFlow | |   | | OpenFlow | |   |
   ||OpenFlow||<....>|  | ------------ |   | ------------ |   |
   |----------|      |  | NE           |   | NE           |   |
   |          |      |  +--------------+   +--------------+   |
   |          |      | Data Plane                             |
   |Management|      +----------------------------------------+

                 Figure 3: High-Level SDN ONF Architecture

   Figure 3 shows the blocks and the functional interfaces of the ONF
   architecture, which comprises three planes: data, controller, and
   application.  The data plane comprehends several Network Entities
   (NEs), which expose their capabilities toward the control plane via a
   Southbound API.  The control plane includes several cooperating
   modules devoted to the creation and maintenance of an abstracted

Bernardos, et al.             Informational                    [Page 10]

RFC 8568       Network Virtualization Research Challenges     April 2019

   resource model of the underlying network.  Such a model is exposed to
   the applications via a Northbound API where the application plane
   comprises several applications/services, each of which has exclusive
   control of a set of exposed resources.

   The management plane spans its functionality across all planes
   performing the initial configuration of the network elements in the
   data plane, the assignment of the SDN controller and the resources
   under its responsibility.  In the control plane, the management needs
   to configure the policies defining the scope of the control given to
   the SDN applications, to monitor the performance of the system and to
   configure the parameters required by the SDN controller modules.  In
   the application plane, the management plane configures the parameters
   of the applications and the service-level agreements.  In addition to
   these interactions, the management plane exposes several functions to
   network operators that can easily and quickly configure and tune the
   network at each layer.

   In RFC 7426 [RFC7426], the IRTF Software-Defined Networking Research
   Group (SDNRG) documented a layer model of an SDN architecture.  This
   was due to the following controversial discussion topics (among
   others).  What exactly is SDN?  What is the layer structure of the
   SDN architecture?  How do layers interface with each other?

   Figure 4 reproduces the figure included in RFC 7426 [RFC7426] to
   summarize the SDN architecture abstractions in the form of a
   detailed, high-level schematic.  In a particular implementation,
   planes can be collocated with other planes or can be physically

   In SDN, a controller manipulates controlled entities via an
   interface.  Interfaces, when local, are mostly API invocations
   through some library or system call.  However, such interfaces may be
   extended via some protocol definition, which may use local
   interprocess communication (IPC) or a protocol that could also act
   remotely; the protocol may be defined as an open standard or in a
   proprietary manner.

   SDN expands multiple planes: forwarding, operational, control,
   management, and application.  All planes mentioned above are
   connected via interfaces.  Additionally, RFC 7426 [RFC7426] considers
   four abstraction layers: the Device and resource Abstraction Layer
   (DAL), the Control Abstraction Layer (CAL), the Management
   Abstraction Layer (MAL), and the Network Services Abstraction Layer

Bernardos, et al.             Informational                    [Page 11]

RFC 8568       Network Virtualization Research Challenges     April 2019

                  |                                |
                  | +-------------+   +----------+ |
                  | | Application |   |  Service | |
                  | +-------------+   +----------+ |
                  |       Application Plane        |
    |           Network Services Abstraction Layer (NSAL)           |
           |                                                |
           |               Service Interface                |
           |                                                |
    o------Y------------------o       o---------------------Y------o
    |      |    Control Plane |       | Management Plane    |      |
    | +----Y----+   +-----+   |       |  +-----+       +----Y----+ |
    | | Service |   | App |   |       |  | App |       | Service | |
    | +----Y----+   +--Y--+   |       |  +--Y--+       +----Y----+ |
    |      |           |      |       |     |               |      |
    | *----Y-----------Y----* |       | *---Y---------------Y----* |
    | | Control Abstraction | |       | | Management Abstraction | |
    | |     Layer (CAL)     | |       | |      Layer (MAL)       | |
    | *----------Y----------* |       | *----------Y-------------* |
    |            |            |       |            |               |
    o------------|------------o       o------------|---------------o
                 |                                 |
                 | CP                              | MP
                 | Southbound                      | Southbound
                 | Interface                       | Interface
                 |                                 |
    |         Device and resource Abstraction Layer (DAL)           |
    |            |                                 |                |
    |    o-------Y----------o   +-----+   o--------Y----------o     |
    |    | Forwarding Plane |   | App |   | Operational Plane |     |
    |    o------------------o   +-----+   o-------------------o     |
    |                       Network Device                          |

                     Figure 4: SDN-Layer Architecture

   While SDN is often directly associated to OpenFlow, this is just one
   (relevant) example of a southbound protocol between the central
   controller and the network entities.  Other relevant examples of
   protocols in the SDN family are NETCONF [RFC6241], RESTCONF
   [RFC8040], and ForCES [RFC5810].

Bernardos, et al.             Informational                    [Page 12]

RFC 8568       Network Virtualization Research Challenges     April 2019

3.3.  ITU-T Functional Architecture of SDN

   The ITU-T (the Telecommunication standardization sector of the
   International Telecommunication Union) has also looked into SDN
   architectures, defining a slightly modified one from what other SDOs
   have done.  In ITU-T recommendation Y.3302 [itu-t-y.3302], the ITU-T
   provides a functional architecture of SDN with descriptions of
   functional components and reference points.  The described functional
   architecture is intended to be used as an enabler for further studies
   on other aspects such as protocols and security as well as being used
   to customize SDN in support of appropriate use cases (e.g., cloud
   computing, mobile networks).  This recommendation is based on ITU-T
   Y.3300 [itu-t-y.3300] and ITU-T Y.3301 [itu-t-y.3301].  While the
   first describes the framework of SDN (including definitions,
   objectives, high-level capabilities, requirements, and the high-level
   architecture of SDN), the second describes more-detailed

   Figure 5 shows the SDN functional architecture defined by the ITU-T.
   It is a layered architecture composed of the SDN application layer
   (SDN-AL), the SDN control layer (SDN-CL), and the SDN resource layer
   (SDN-RL).  It also has multi-layer management functions (MMF), which
   provide the ability to manage the functionalities of SDN layers,
   i.e., SDN-AL, SDN-CL, and SDN-RL.  MMF interacts with these layers
   using Multi-layer Management Functions Application (MMFA), Multi-
   layer Management Functions Control (MMFC), and Multi-layer Management
   Functions Resource (MMFR) reference points.

   The SDN-AL enables a service-aware behavior of the underlying network
   in a programmatic manner.  The SDN-CL provides programmable means to
   control the behavior of SDN-RL resources (such as data transport and
   processing) following requests received from the SDN-AL according to
   MMF policies.  The SDN-RL is where the physical or virtual network
   elements perform transport and/or processing of data packets
   according to SDN-CL decisions.

Bernardos, et al.             Informational                    [Page 13]

RFC 8568       Network Virtualization Research Challenges     April 2019

          MMFO                      MMFA
   +-----+ . +---------------------+ . +--------------------+
   |     | . |+---+ +---+ +-------+| . |+---------+ +-----+ |
   |     | . ||   | |   | |       || . ||   AL.   | |     | |
   |     | . || E | |   | |  App. || . || Mngmt.  | | SDN | | SDN-AL
   |     | . || x | | M | | Layer || . || Support | | App | |
   |     | . || t.| | u | | Mngmt.|| . || & Orch. | |     | |
   |     | . ||   | | l | +-------+| . |+---------+ +-----+ |
   |     | . || R | | t |          | . +--------------------+
   |     | . || e | | i |          |MMFC ..................... ACI
   |     | . || l | | - |          | . +--------------------+
   |     | . || a | | l | +-------+| . |+------+ +---------+|
   | OSS/| . || t | | a | |       || . ||      | |   App.  ||
   | BSS | . || i | | y | |       || . ||      | | Support ||
   |     | . || o | | e | |       || . ||      | +---------+|
   |     | . || n | | r | |       || . ||  CL  | +---------+|
   |     | . || s | |   | |Control|| . ||Mngmt.| | Control ||
   |     | . || h | | M | | Layer || . || Supp.| |  Layer  || SDN-CL
   |     | . || i | | a | | Mngmt.|| . || and  | |  Serv.  ||
   |     | . || p | | n | |       || . || Orch.| +---------+|
   |     | . ||   | | a | |       || . ||      | +---------+|
   |     | . || M | | g | |       || . ||      | | Resource||
   |     | . || n | | e | |       || . ||      | | Abstrac.||
   |     | . || g | | m | +-------+| . |+------+ +---------+|
   |     | . || m | | e |          | . +--------------------+
   |     | . || t.| | n |          |MMFR ..................... RCI
   |     | . ||   | | t |          | . +--------------------+
   +-----+ . |+---+ |   | +-------+| . |+------++----------+|
             |      | O | |       || . ||      ||RL Control||
             |      | r | |Resour.|| . ||  RL  |+----------+|
        MMF  |      | c | | Layer || . ||Mngmt.|+----++----+| SDN-RL
             |      | h.| | Mngmt.|| . || Supp.||Data||Data||
             |      |   | |       || . ||      ||Tran||Proc||
             |      +---+ +-------+| . |+------++----++----+|
             +---------------------+ . +--------------------+

     ACI:  Application Control Interface
     MMFA: Multi-layer Management Functions Application
     MMFC: Multi-layer Management Functions Control
     MMFO: Multi-layer Management Functions OSS/BSS
     MMFR: Multi-layer Management Functions Resource
     RCI:  Resource Control Interfaces
     RL:   Resource Layer

                Figure 5: ITU-T SDN Functional Architecture

Bernardos, et al.             Informational                    [Page 14]

RFC 8568       Network Virtualization Research Challenges     April 2019

3.4.  Multi-Access Edge Computing

   Multi-access Edge Computing (MEC) -- formerly known as Mobile Edge
   Computing -- capabilities deployed in the edge of the mobile network
   can facilitate the efficient and dynamic provision of services to
   mobile users.  The ETSI ISG MEC working group, operative from end of
   2014, intends to specify an open environment for integrating MEC
   capabilities with service providers' networks, also including
   applications from third parties.  These distributed computing
   capabilities provide IT infrastructure as in a cloud environment for
   the deployment of functions in mobile access networks.  It can be
   seen then as a complement to both NFV and SDN.

3.5.  IEEE 802.1CF (OmniRAN)

   The IEEE 802.1CF Recommended Practice [omniran] specifies an access
   network that connects terminals to their access routers utilizing
   technologies based on the family of IEEE 802 Standards (e.g., 802.3
   Ethernet, 802.11 Wi-Fi, etc.).  The specification defines an access
   network reference model, including entities and reference points
   along with behavioral and functional descriptions of communications
   among those entities.

   The goal of this project is to help unify the support of different
   interfaces, enabling shared-network control and use of SDN
   principles, thereby lowering the barriers to new network
   technologies, to new network operators, and to new service providers.

3.6.  Distributed Management Task Force (DMTF)

   The DMTF <https://www.dmtf.org/> is an industry standards
   organization working to simplify the manageability of network-
   accessible technologies through open and collaborative efforts by
   some technology companies.  The DMTF is involved in the creation and
   adoption of interoperable management standards, supporting
   implementations that enable the management of diverse traditional and
   emerging technologies including cloud, virtualization, network, and

   There are several DMTF initiatives that are relevant to the network
   virtualization area, such as the Open Virtualization Format (OVF) for
   VNF packaging; the Cloud Infrastructure Management Interface (CIMI)
   for cloud infrastructure management; the Network Management (NETMAN),
   for VNF management; and the Virtualization Management (VMAN), for
   virtualization infrastructure management.

Bernardos, et al.             Informational                    [Page 15]

RFC 8568       Network Virtualization Research Challenges     April 2019

3.7.  Open-Source Initiatives

   The open-source community is especially active in the area of network
   virtualization and orchestration.  We next summarize some of the
   active efforts:

   o  OpenStack.  OpenStack is a free and open-source cloud-computing
      software platform.  OpenStack software controls large pools of
      compute, storage, and networking resources throughout a data
      center, managed through a dashboard or via the OpenStack API.

   o  Kubernetes.  Kubernetes is an open-source system for automating
      deployment, scaling and management of containerized applications.
      Kubernetes can schedule and run application containers on clusters
      of physical or virtual machines.  Kubernetes allows (i) Scale on
      the fly, (ii) Limit hardware usage to required resources only,
      (iii) Load-balancing Monitoring, and (iv) Efficient life-cycle

   o  OpenDayLight.  OpenDayLight (ODL) is a highly available, modular,
      extensible and scalable multiprotocol controller infrastructure
      built for SDN deployments on modern heterogeneous multi-vendor
      networks.  It provides a model-driven service abstraction platform
      that allows users to write apps that easily work across a wide
      variety of hardware and southbound protocols.

   o  ONOS.  The Open Network Operating System (ONOS) project is an
      open-source community hosted by The Linux Foundation.  The goal of
      the project is to create an SDN operating system for
      communications service providers that is designed for scalability,
      high performance, and high availability.

   o  OpenContrail.  OpenContrail is a licensed Apache 2.0 project that
      is built using standards-based protocols and that provides all the
      necessary components for network virtualization: an SDN
      controller, a virtual router, an analytics engine, and published
      northbound APIs.  It has an extensive Representational State
      Transfer (REST) API to configure and gather operational and
      analytics data from the system.

   o  OPNFV.  The Open Platform for NFV (OPNFV) is a carrier-grade,
      integrated, open-source platform to accelerate the introduction of
      new NFV products and services.  By integrating components from
      upstream projects, the OPNFV community aims at conducting
      performance and use case-based testing to ensure the platform's
      suitability for NFV use cases.  The scope of OPNFV's initial
      release is focused on building NFV Infrastructure (NFVI) and
      Virtualized Infrastructure Manager (VIM) by integrating components

Bernardos, et al.             Informational                    [Page 16]

RFC 8568       Network Virtualization Research Challenges     April 2019

      from upstream projects such as OpenDayLight, OpenStack, Ceph
      Storage, Kernel-based Virtual Machine (KVM), Open vSwitch, and
      Linux.  These components, along with APIs to other NFV elements,
      form the basic infrastructure required for Virtualized Network
      Functions (VNFs) and Management and Orchestration (MANO)
      components.  OPNFV's goal is to (i) increase performance and power
      efficiency, (ii) improve reliability, availability, and
      serviceability, and (iii) deliver comprehensive platform

   o  OSM.  Open Source Mano (OSM) is an ETSI-hosted project to develop
      an Open Source NFV Management and Orchestration (MANO) software
      stack aligned with ETSI NFV.  OSM is based on components from
      previous projects, such Telefonica's OpenMANO or Canonical's Juju,
      among others.

   o  OpenBaton.  OpenBaton is a Network Function Virtualization
      Orchestrator (NFVO) that is ETSI NFV compliant.  OpenBaton was
      part of the OpenSDNCore project started with the objective of
      providing a compliant implementation of the ETSI NFV

   o  ONAP.  Open Network Automation Platform (ONAP) is an open-source
      software platform that delivers capabilities for the design,
      creation, orchestration, monitoring, and life-cycle management of
      (i) Virtual Network Functions (VNFs), (ii) The carrier-scale
      Software-Defined Networks (SDNs) that contain them, and (iii)
      higher-level services that combine the above.  ONAP (derived from
      the AT&T's ECOMP) provides for automatic, policy-driven
      interaction of these functions and services in a dynamic, real-
      time cloud environment.

   o  SONA.  The Simplified Overlay Network Architecture (SONA) is an
      extension to ONOS to have an almost full SDN network control in
      OpenStack for virtual tenant network provisioning.  Basically,
      SONA is an SDN-based network virtualization solution for cloud DC.

   Among the main areas that are being developed by the aforementioned
   open-source activities that relate to network virtualization
   research, we can highlight policy-based resource management,
   analytics for visibility and orchestration, and service verification
   with regard to security and resiliency.

Bernardos, et al.             Informational                    [Page 17]

RFC 8568       Network Virtualization Research Challenges     April 2019

4.  Network Virtualization Challenges

4.1.  Overview

   Network virtualization is changing the way the telecommunications
   sector will deploy, extend, and operate their networks.  These new
   technologies aim at reducing the overall costs by moving
   communication services from specific hardware in the operators' cores
   to server farms scattered in data centers (i.e., compute and storage
   virtualization).  In addition, the networks interconnecting the
   functions that compose a network service are fundamentally affected
   in the way they route, process, and control traffic (i.e., network

4.2.  Guaranteeing Quality of Service

   Achieving a given QoS in an NFV environment with virtualized and
   distributed computing, storage, and networking functions is more
   challenging than providing the equivalent in discrete non-virtualized
   components.  For example, ensuring a guaranteed and stable forwarding
   data rate has proven not to be straightforward when the forwarding
   function is virtualized and runs on top of COTS server hardware
   [openmano_dataplane] [NFV-COTS] [etsi_nfv_whitepaper_3].  Again, the
   comparison point is against a router or forwarder built on optimized
   hardware.  We next identify some of the challenges that this poses.

4.2.1.  Virtualization Technologies

   The issue of guaranteeing a network QoS is less of an issue for
   "traditional" cloud computing because the workloads that are treated
   there are servers or clients in the networking sense and hardly ever
   process packets.  Cloud computing provides hosting for applications
   on shared servers in a highly separated way.  Its main advantage is
   that the infrastructure costs are shared among tenants and that the
   cloud infrastructure provides levels of reliability that can not be
   achieved on individual premises in a cost-efficient way
   [intel_10_differences_nfv_cloud].  NFV has very strict requirements
   posed in terms of performance, stability, and consistency.  Although
   there are some tools and mechanisms to improve this, such as Enhanced
   Performance Awareness (EPA), Single Root I/O Virtualization (SR-IOV),
   Non-Uniform Memory Access (NUMA), Data Plane Development Kit (DPDK),
   etc., these are still unsolved challenges.  One open research issue
   is finding out technologies that are different from Virtual Machines
   (VMs) and more suitable for dealing with network functionalities.

   Lately, a number of lightweight virtualization technologies including
   containers, unikernels (specialized VMs) and minimalistic
   distributions of general-purpose OSes have appeared as virtualization

Bernardos, et al.             Informational                    [Page 18]

RFC 8568       Network Virtualization Research Challenges     April 2019

   approaches that can be used when constructing an NFV platform.
   [LIGHT-NFV] describes the challenges in building such a platform and
   discusses to what extent these technologies, as well as traditional
   VMs, are able to address them.

4.2.2.  Metrics for NFV Characterization

   Another relevant aspect is the need for tools for diagnostics and
   measurements suited for NFV.  There is a pressing need to define
   metrics and associated protocols to measure the performance of NFV.
   Specifically, since NFV is based on the concept of taking centralized
   functions and evolving them to highly distributed software (SW)
   functions, there is a commensurate need to fully understand and
   measure the baseline performance of such systems.

   The IP Performance Metrics (IPPM) WG defines metrics that can be used
   to measure the quality and performance of Internet services and
   applications running over transport-layer protocols (e.g., TCP and
   UDP) over IP.  It also develops and maintains protocols for the
   measurement of these metrics.  While the IPPM WG is a long-running WG
   that started in 1997, at the time of writing, it does not have a
   charter item or active Internet-Drafts related to the topic of
   network virtualization.  In addition to using IPPM to evaluate QoS,
   there is a need for specific metrics for assessing the performance of
   network-virtualization techniques.

   The Benchmarking Methodology Working Group (BMWG) is also performing
   work related to NFV metrics.  For example, [RFC8172] investigates
   additional methodological considerations necessary when benchmarking
   VNFs that are instantiated and hosted in general-purpose hardware,
   using bare-metal hypervisors or other isolation environments (such as
   Linux containers).  An essential consideration is benchmarking
   physical and VNFs in the same way when possible, thereby allowing
   direct comparison.

   There is a clear motivation for the work on performance metrics for
   NFV [etsi_gs_nfv_per_001], as stated in [RFC8172] (and replicated

      I'm designing and building my NFV Infrastructure platform.  The
      first steps were easy because I had a small number of categories
      of VNFs to support and the VNF vendor gave HW recommendations that
      I followed.  Now I need to deploy more VNFs from new vendors, and
      there are different hardware recommendations.  How well will the
      new VNFs perform on my existing hardware?  Which among several new
      VNFs in a given category are most efficient in terms of capacity
      they deliver?  And, when I operate multiple categories of VNFs
      (and PNFs) *concurrently* on a hardware platform such that they

Bernardos, et al.             Informational                    [Page 19]

RFC 8568       Network Virtualization Research Challenges     April 2019

      share resources, what are the new performance limits, and what are
      the software design choices I can make to optimize my chosen
      hardware platform?  Conversely, what hardware platform upgrades
      should I pursue to increase the capacity of these concurrently
      operating VNFs?

   Lately, there are also some efforts looking into VNF benchmarking.
   The selection of an NFV Infrastructure Point of Presence to host a
   VNF or allocation of resources (e.g., virtual CPUs, memory) needs to
   be done over virtualized (abstracted and simplified) resource views
   [vnf_benchmarking] [VNF-VBAAS].

4.2.3.  Predictive Analysis

   On top of diagnostic tools that enable an assessment of the QoS,
   predictive analyses are required to react before anomalies occur.
   Due to the SW characteristics of VNFs, a reliable diagnosis framework
   could potentially enable the prevention of issues by a proper
   diagnosis and then a reaction in terms of acting on the potentially
   impacted service (e.g., migration to a different compute node,
   scaling in/out, up/down, etc.).

4.2.4.  Portability

   Portability in NFV refers to the ability to run a given VNF on
   multiple NFVIs, that is, guaranteeing that the VNF would be able to
   perform its functions with a high and predictable performance given
   that a set of requirements on the NFVI resources is met.  Therefore,
   portability is a key feature that, if fully enabled, would contribute
   to making the NFV environment achieve a better reliability than a
   traditional system.  Implementing functionality in SW over
   "commodity" infrastructure should make it much easier to port/move
   functions from one place to another.  However, this is not yet as
   ideal as it sounds, and there are aspects that are not fully tackled.
   The existence of different hypervisors, specific hardware
   dependencies (e.g., EPA related), or state-synchronization aspects
   are just some examples of troublemakers for portability purposes.

   The ETSI NFV ISG is doing work in relation to portability.
   [etsi_gs_nfv_per_001] provides a list of minimal features that the VM
   Descriptor and Compute Host Descriptor should contain for the
   appropriate deployment of VM images over an NFVI (i.e., a "telco data
   center"), in order to guarantee high and predictable performance of
   data-plane workloads while assuring their portability.  In addition,
   [etsi_gs_nfv_per_001] provides a set of recommendations on the
   minimum requirements that hardware (HW) and hypervisor should have
   for a "telco data center" suitable for different workloads (data
   plane, control plane, etc.) present in VNFs.  The purpose of

Bernardos, et al.             Informational                    [Page 20]

RFC 8568       Network Virtualization Research Challenges     April 2019

   [etsi_gs_nfv_per_001] is to provide the list of VM requirements that
   should be included in the VM Descriptor template, and the list of HW
   capabilities that should be included in the Compute Host Descriptor
   (CHD) to assure predictable high performance.  ETSI NFV assumes that
   the MANO functions will make the mix & match.  Therefore, there are
   still several research challenges to be addressed here.

4.3.  Performance Improvement

4.3.1.  Energy Efficiency

   Virtualization is typically seen as a direct enabler of energy
   savings.  Some of the enablers for this that are often mentioned
   [nfv_sota_research_challenges] are (i) the multiplexing gains
   achieved by centralizing functions in data centers reduce the overall
   energy consumed and (ii) the flexibility brought by network
   programmability enables to switch off infrastructure as needed in a
   much easier way.  However, there is still a lot of room for
   improvement in terms of virtualization techniques to reduce the power
   consumption, such as enhanced-hypervisor technologies.

   Some additional examples of research topics that could enable energy
   savings are [nfv_sota_research_challenges]:

   o  Energy-aware scaling (e.g., reductions in CPU speeds and partially
      turning off some hardware components to meet a given energy
      consumption target.

   o  Energy-aware function placement.

   o  Scheduling and chaining algorithms, for example, adapting the
      network topology and operating parameters to minimize the
      operation cost (e.g., tracking energy costs to identify the
      cheapest prices).

   Note that it is also important to analyze the trade-off between
   energy efficiency and network performance.

4.3.2.  Improved Link Usage

   The use of NFV and SDN technologies can help improve link usage.  SDN
   has already shown that it can greatly increase average link
   utilization (e.g., Google example [google_sdn_wan]).  NFV adds more
   complexity (e.g., due to service-function chaining / VNF forwarding
   graphs), which needs to be considered.  Aspects like the ones
   described in [NFVRG-TOPO] (on NFV data center topology design) have
   to be looked at carefully as well.

Bernardos, et al.             Informational                    [Page 21]

RFC 8568       Network Virtualization Research Challenges     April 2019

4.4.  Multiple Domains

   Market fragmentation has resulted in a multitude of network operators
   each focused on different countries and regions.  This makes it
   difficult to create infrastructure services spanning multiple
   countries, such as virtual connectivity or compute resources, as no
   single operator has a footprint everywhere.  Cross-domain
   orchestration of services over multiple administrations or over
   multi-domain single administrations will allow end-to-end network and
   service elements to mix in multi-vendor, heterogeneous technology,
   and resource environments [multi-domain_5GEx].

   For the specific use case of 'Network as a Service', it becomes even
   more important to ensure that Cross Domain Orchestration also takes
   care of hierarchy of networks and their association, with respect to
   provisioning tunnels and overlays.

   Multi-domain orchestration is currently an active research topic,
   which is being tackled, among others, by ETSI NFV ISG and the 5GEx
   project <https://www.5gex.eu/> [MULTI-NMRG] [multi-domain_5GEx].

   Another side of the multi-domain problem is the integration/
   harmonization of different management domains.  A key example comes
   from Multi-access Edge Computing, which, according to ETSI, comes
   with its own MANO system and would require integration if
   interconnected to a generic NFV system.

4.5.  5G and Network Slicing

   From the beginning of all 5G discussions in the research and industry
   fora, it has been agreed that 5G will have to address many more use
   cases than the preceding wireless generations, which first focused on
   voice services and then on voice and high-speed packet data services.
   In this case, 5G should be able to handle not only the same (or
   enhanced) voice and packet data services, but also emerging services
   like tactile Internet and the Internet of Things (IoT).  These use
   cases take the requirements to opposite extremes, as some of them
   require ultra-low latency and higher-speed, whereas some others
   require ultra-low power consumption and high-delay tolerance.

   Because of these very extreme 5G use cases, it is envisioned that
   selective combinations of radio access networks and core network
   components will have to be combined into a given network slice to
   address the specific requirements of each use case.

   For example, within the major IoT category, which is perhaps the most
   disrupting one, some autonomous IoT devices will have very low
   throughput, will have much longer sleep cycles (and therefore high

Bernardos, et al.             Informational                    [Page 22]

RFC 8568       Network Virtualization Research Challenges     April 2019

   latency), and a battery life time exceeding by a factor of thousands
   that of smartphones or some other devices that will have almost
   continuous control and data communications.  Hence, it is envisioned
   that a customized network slice will have to be stitched together
   from virtual resources or sub-slices to meet these requirements.

   The actual definition of a "network slice" from an IP infrastructure
   viewpoint is currently undergoing intense debate; see [COMS-PS],
   [NETSLICES], [SLICE-3GPP], and [ngmn_5G_whitepaper].  Network slicing
   is a key for introducing new actors in existing markets at a low cost
   -- by letting new players rent "blocks" of capacity, if the new
   business model enables performance that meets the application needs
   (e.g., broadcasting updates to many sensors with satellite
   broadcasting capabilities).  However, more work needs to be done to
   define the basic architectural approach of how network slices will be
   defined and formed.  For example, is it mostly a matter of defining
   the appropriate network models (e.g., YANG) to stitch the network
   slice from existing components?  Or do end-to-end timing,
   synchronization, and other low-level requirements mean that more
   fundamental research has to be done?

4.5.1.  Virtual Network Operators

   The widespread use/discussion/practice of system and network
   virtualization technologies has led to new business opportunities,
   enlarging the offer of IT resources with virtual network and
   computing resources, among others.  As a consequence, the network
   ecosystem now differentiates between the owner of physical resources,
   the Infrastructure Provider (InP), and the intermediary that conforms
   and delivers network services to the final customers, the Virtual
   Network Operator (VNO).

   VNOs aim to exploit the virtualized infrastructures to deliver new-
   and-improved services to their customers.  However, current network
   virtualization techniques offer poor support for VNOs to control
   their resources.  It has been considered that the InP is responsible
   for the reliability of the virtual resources, but there are several
   situations in which a VNO requires a finer control on its resources.
   For instance, dynamic events, such as the identification of new
   requirements or the detection of incidents within the virtual system,
   might urge a VNO to quickly reform its virtual infrastructure and
   resource allocation.  However, the interfaces offered by current
   virtualization platforms do not offer the necessary functions for
   VNOs to perform the elastic adaptations they need to conduct in
   dynamic environments.

Bernardos, et al.             Informational                    [Page 23]

RFC 8568       Network Virtualization Research Challenges     April 2019

   Beyond their heterogeneity, which can be resolved by software
   adapters, current virtualization platforms do not have common methods
   and functions, so it is difficult for the virtual network controllers
   used by the VNOs to actually manage and control virtual resources
   instantiated on different platforms, not even considering different
   InPs.  Therefore, it is necessary to reach a common definition of the
   functions that should be offered by underlying platforms to give such
   overlay controllers the possibility to allocate and deallocate
   resources dynamically and get monitoring data about them.

   Such common methods should be offered by all underlying controllers,
   regardless of being network-oriented (e.g., ODL, ONOS, and Ryu) or
   computing-oriented (e.g., OpenStack, OpenNebula, and Eucalyptus).
   Furthermore, it is important for those platforms to offer some "PUSH"
   function to report resource state, avoiding the need for the VNO's
   controller to "POLL" for such data.  A starting point to get proper
   notifications within current REST APIs could be to consider the
   protocol proposed by the WEBPUSH WG [RFC8030].

   Finally, in order to establish a proper order and allow the
   coexistence and collaboration of different systems, a common ontology
   regarding network and system virtualization should be defined and
   agreed upon, so different and heterogeneous systems can understand
   each other without requiring reliance on specific adaptation
   mechanisms that might break with any update on any side of the

4.5.2.  Extending Virtual Networks and Systems to the Internet of Things

   The Internet of Things (IoT) refers to the vision of connecting a
   multitude of automated devices (e.g., lights, environmental sensors,
   traffic lights, parking meters, health and security systems, etc.) to
   the Internet for purposes of reporting and remote command and control
   of the device.  This vision is being realized by a multi-pronged
   approach of standardization in various forums and complementary open-
   source activities.  For example, in the IETF, support of IoT web
   services has been defined by an HTTP-like protocol adapted for IoT
   called "CoAP" [RFC7252]; and, lately, a group has been studying the
   need to develop a new network layer to support IP applications over
   Low-Power Wide Area Networks (LPWAN).

   Elsewhere, for 5G cellular evolution, there is much discussion on the
   need for supporting virtual network slices for the expected massive
   numbers of IoT devices.  A separate virtual network slice is
   considered necessary for different 5G IoT use cases because devices
   will have very different characteristics than typical cellular

Bernardos, et al.             Informational                    [Page 24]

RFC 8568       Network Virtualization Research Challenges     April 2019

   devices like smartphones [ngmn_5G_whitepaper], and the number of IoT
   devices is expected to be at least one or two orders of magnitude
   higher than other 5G devices (see Section 4.5).

   The specific nature of the IoT ecosystem, particularly reflected in
   the Machine-to-Machine (M2M) communications, leads to the creation of
   new and highly distributed systems which demand location-based
   network and computing services.  A specific example can be
   represented by a set of "things" that suddenly require the setup of a
   firewall to allow external entities to access their data while
   outsourcing some computation requirements to more powerful systems
   relying on cloud-based services.  This representative use case
   exposes important requirements for both NFV and the underlying cloud

   In order to provide the aforementioned location-based functions
   integrated with highly distributed systems, the so-called fog
   infrastructures should be able to instantiate VNFs, placing them in
   the required place, e.g., close to their consumers.  This requirement
   implies that the interfaces offered by virtualization platforms must
   support the specification of location-based resources, which is a key
   function in those scenarios.  Moreover, those platforms must also be
   able to interpret and understand the references used by IoT systems
   to their location (e.g., "My-AP" or "5BLDG+2F") and also the
   specification of identifiers linked to other resources, such as the
   case of requiring the infrastructure to establish a link between a
   specific Access Point (AP) and a specific virtual computing node.  In
   summary, the research gap is exact localization of VNFs at far
   network edge infrastructure, which is highly distributed and dynamic.

4.6.  Service Composition

   Current network services deployed by operators often involve the
   composition of several individual functions (such as packet
   filtering, deep-packet inspection, load-balancing).  These services
   are typically implemented by the ordered combination of a number of
   service functions that are deployed at different points within a
   network, not necessarily on the direct data path.  This requires
   traffic to be steered through the required service functions,
   wherever they are deployed [RFC7498].

   For a given service, the abstracted view of the required service
   functions and the order in which they are to be applied is called
   "Service Function Chaining" (SFC) [sfc_challenges], which is called
   "Network Function Forwarding Graph" (NF-FG) in ETSI.  SFC is
   instantiated through the selection of specific service function
   instances on specific network nodes to form a service graph: this is

Bernardos, et al.             Informational                    [Page 25]

RFC 8568       Network Virtualization Research Challenges     April 2019

   called a "Service Function Path" (SFP).  The service functions may be
   applied at any layer within the network protocol stack (network
   layer, transport layer, application layer, etc.).

   Service composition is a powerful means that can provide significant
   benefits when applied in a softwarized network environment.  However,
   there are many research challenges in this area; for example, the
   ones related to composition mechanisms and algorithms to enable load-
   balancing and improve reliability.  The service composition should
   also act as an enabler to gather information across all hierarchies
   (underlays and overlays) of network deployments that may span across
   multiple operators for faster serviceability, thus facilitating
   accomplishing aforementioned goals of "load-balancing and improving

   As described in [dynamic_chaining], different algorithms can be used
   to enable dynamic service composition that optimizes a QoS-based
   utility function (e.g., minimizing the latency per-application
   traffic flows) for a given composition plan.  Such algorithms can
   consider the computation capabilities and load status of resources
   executing the VNF instances, either deduced through estimations from
   historical usage data or collected through real-time monitoring
   (i.e., context-aware selection).  For this reason, selections should
   include references to dynamic information on the status of the
   service instance and its constituent elements, i.e., monitoring
   information related to individual VNF instances and links connecting
   them as well as derived monitoring information at the chain level
   (e.g., end-to-end delay).  At runtime, if one or more VNF instances
   are no longer available or QoS degrades below a given threshold, the
   service selection task can be rerun to perform service substitution.

   There are different research directions that relate to the previous
   point.  For example, the use of Integer Linear Programming (ILP)
   techniques can be explored to optimize the management of diverse
   traffic flows.  Deep-machine learning can also be applied to optimize
   service chains using information parameters, such as some of the ones
   mentioned above.  Newer scheduling paradigms, like co-flows, can also
   be used.

   The SFC working group is working on an architecture for SFC [RFC7665]
   that includes the necessary protocols or protocol extensions to
   convey the SFC and SFP information to nodes that are involved in the
   implementation of service functions and SFCs as well as mechanisms
   for steering traffic through service functions.

   In terms of actual work items, the SFC WG has not yet considered
   working on the management and configuration of SFC components related
   to the support of SFC.  This part is of special interest for

Bernardos, et al.             Informational                    [Page 26]

RFC 8568       Network Virtualization Research Challenges     April 2019

   operators and would be required in order to actually put SFC
   mechanisms into operation.  Similarly, redundancy and reliability
   mechanisms for SFC are currently not dealt with by any WG in the
   IETF.  While this was the main goal of the VNFpool BoF efforts, it
   still remains unaddressed.

4.7.  Device Virtualization for End Users

   So far, most of the network softwarization efforts have focused on
   virtualizing functions of network elements.  While virtualization of
   network elements started with the core, mobile-network architectures
   are now heavily switching to also virtualize Radio Access Network
   (RAN) functions.  The next natural step is to get virtualization down
   at the level of the end-user device (e.g., virtualizing a smartphone)
   [virtualization_mobile_device].  The cloning of a device in the cloud
   (central or local) bears attractive benefits to both the device and
   network operations alike (e.g., power saving at the device by
   offloading computational-heaving functions to the cloud, optimized
   networking -- both device-to-device and device-to-infrastructure) for
   service delivery through tighter integration of the device (via its
   clone in the networking infrastructure).  This is, for example, being
   explored by the European H2020 ICIRRUS project

4.8.  Security and Privacy

   Similar to any other situations where resources are shared, security
   and privacy are two important aspects that need to be taken into

   In the case of security, there are situations where multiple service
   providers will need to coexist in a virtual or hybrid physical/
   virtual environment.  This requires attestation procedures amongst
   different virtual/physical functions and resources as well as ongoing
   external monitoring.  Similarly, different network slices operating
   on the same infrastructure can present security problems, for
   instance, if one slice running critical applications (e.g., support
   for a safety system) is affected by another slice running a less
   critical application.  In general, the minimum common denominator for
   security measures on a shared system should be equal to or higher
   than the one required by the most-critical application.  Multiple and
   continuous threat model analysis as well as a DevOps model are
   required to maintain a certain level of security in an NFV system.
   Simplistically, DevOps is a process that combines multiple functions
   into single cohesive teams in order to quickly produce quality
   software.  Typically, it relies on also applying the Agile
   development process, which focuses on (among many things) dividing
   large features into multiple, smaller deliveries.  One part of this

Bernardos, et al.             Informational                    [Page 27]

RFC 8568       Network Virtualization Research Challenges     April 2019

   is to immediately test the new smaller features in order to get
   immediate feedback on errors so that if present, they can be
   immediately fixed and redeployed.

   On the other hand, privacy refers to concerns about the control of
   personal data and the decision of what to reveal to whom.  In this
   case, the storage, transmission, collection, and potential
   correlation of information in the NFV system, for purposes not
   originally intended or not known by the user, should be avoided.
   This is particularly challenging, as future intentions and threats
   cannot be easily predicted and still can be applied on data collected
   in the past.  Therefore, well-known techniques, such as data
   minimization using privacy features as default and allowing users to
   opt in/out, should be used to prevent potential privacy issues.

   Compared to traditional networks, NFV will result in networks that
   are much more dynamic (in function distribution and topology) and
   elastic (in size and boundaries).  Thus, NFV will require network
   operators to evolve their operational and administrative security
   solutions to work in this new environment.  For example, in NFV, the
   network orchestrator will become a key node to provide security
   policy orchestration across the different physical and virtual
   components of the virtualized network.  For highly confidential data,
   for example, the network orchestrator should take into account if
   certain physical HW of the network is considered to be more secure
   (e.g., because it is located in secure premises) than other HW.

   Traditional telecom networks typically run under a single
   administrative domain controlled by (exactly) one operator.  With
   NFV, it is expected that in many cases, the telecom operator will now
   become a tenant (running the VNFs), and the infrastructure (NFVI) may
   be run by a different operator and/or cloud service provider (see
   also Section 4.4).  Thus, there will be multiple administrative
   domains involved, making security policy coordination more complex.
   For example, who will be in charge of provisioning and maintaining
   security credentials such as public and private keys?  Also, should
   private keys be allowed to be replicated across the NFV for
   redundancy reasons?  Alternatively, it can be investigated how to
   develop a mechanism that avoids such a security policy coordination,
   thus making the system more robust.

   On a positive note, NFV may better defend against denial-of-service
   (DoS) attacks because of the distributed nature of the network (i.e.,
   no single point of failure) and the ability to steer (undesirable)
   traffic quickly [etsi_gs_nfv_sec_001].  Also, NFVs that have physical
   HW that is distributed across multiple data centers will also provide

Bernardos, et al.             Informational                    [Page 28]

RFC 8568       Network Virtualization Research Challenges     April 2019

   better fault isolation environments.  Particularly, this holds true
   if each data center is protected separately via firewalls,
   Demilitarized Zones (DMZs), and other network-protection techniques.

   SDN can also be used to help improve security by facilitating the
   operation of existing protocols, such as Authentication,
   Authorization and Accounting (AAA).  The management of AAA
   infrastructures, namely the management of AAA routing and the
   establishment of security associations between AAA entities, can be
   performed using SDN, as analyzed in [SDN-AAA].

4.9.  Separation of Control Concerns

   NFV environments offer two possible levels of SDN control.  One level
   is the need for controlling the NFVI to provide connectivity end-to-
   end among VNFs or among VNFs and Physical Network Functions (PNFs).
   A second level is the control and configuration of the VNFs
   themselves (in other words, the configuration of the network service
   implemented by those VNFs), taking advantage of the programmability
   brought by SDN.  Both control concerns are separated in nature.
   However, interaction between both could be expected in order to
   optimize, scale, or influence each other.

   Clear mechanisms for such interactions are needed in order to avoid
   malfunctioning or interference concerns.  These ideas are considered
   in [etsi_gs_nfv_eve005] and [LAYERED-SDN].

4.10.  Network Function Placement

   Network function placement is a problem in any kind of network
   telecommunications infrastructure.  Moreover, the increased degree of
   freedom added by network virtualization makes this problem even more
   important, and also harder to tackle.  Deciding where to place VNFs
   is a resource-allocation problem that needs to (or may) take into
   consideration quite a few aspects: resiliency, (anti-)affinity,
   security, privacy, energy efficiency, etc.

   When several functions are chained (typical scenario), placement
   algorithms become more complex and important (as described in
   Section 4.6).  While there has been research on the topic
   ([nfv_piecing], [dynamic_placement], and [vnf-p]), this still remains
   an open challenge that requires more attention.  The use of multi-
   domains adds another component of complexity to this problem that has
   to be considered.

Bernardos, et al.             Informational                    [Page 29]

RFC 8568       Network Virtualization Research Challenges     April 2019

4.11.  Testing

   The impacts of network virtualization on testing can be divided into
   three groups:

   1.  Changes in methodology

   2.  New functionality

   3.  Opportunities

4.11.1.  Changes in Methodology

   The largest impact of NFV is the ability to isolate the System Under
   Test (SUT).  When testing PNFs, isolating the SUT means that all the
   other devices that the SUT communicates with are replaced with
   simulations (or controlled executions) in order to place the SUT
   under test by itself.  The SUT may be comprised of one or more
   devices.  The simulations use the appropriate traffic type and
   protocols in order to execute test cases.

   As shown in Figure 2, NFV provides a common architecture for all
   functions to use.  A VNF is executed using resources offered by the
   NFVI, which have been allocated using the MANO function.  It is not
   possible to test a VNF by itself, without the entire supporting
   environment present.  This fundamentally changes how to consider the
   SUT.  In the case of a VNF (or multiple VNFs), the SUT is part of a
   larger architecture that is necessary in order to run the SUTs.

   Therefore, isolation of the SUT becomes controlling the environment
   in a disciplined manner.  The components of the environment necessary
   to run the SUTs that are not part of the SUT itself become the test
   environment.  In the case of VNFs that are part of the SUT, the NFVI
   and MANO become the test environment.  The configurations and
   policies that guide the test environment should remain constant
   during the execution of the tests, and also from test to test.
   Configurations such as CPU pinning, NUMA configuration, the SW
   versions and configurations of the hypervisor, vSwitch and NICs
   should remain constant.  The only variables in the testing should be
   those controlling the SUT itself.  If any configuration in the test
   environment is changed from test to test, the results become very
   difficult, if not impossible, to compare since the test environment
   behavior may change the results as a consequence of the configuration

   Testing the NFVI itself also presents new considerations.  With a
   PNF, the dedicated hardware supporting it is optimized for the
   particular workload of the function.  Routing hardware is specially

Bernardos, et al.             Informational                    [Page 30]

RFC 8568       Network Virtualization Research Challenges     April 2019

   built to support packet forwarding functions, while the hardware to
   support a purely control-plane application (say, a DNS server, or a
   Diameter function) will not have this specialized capability.  In
   NFV, the NFVI is required to support all types of potentially
   different workload types.

   Therefore, testing the NFVI requires careful consideration about what
   types of metrics are sought.  This, in turn, depends on the workload
   type the expected VNF will be.  Examples of different workload types
   are data forwarding, control plane, encryption, and authentication.
   All these types of expected workloads will determine the types of
   metrics that should be sought.  For example, if the workload is
   control plane, then a metric such as jitter is not useful, but
   dropped packets are critical.  In a multi-tenant environment, the
   NFVI could support various types of workloads.  In this case, testing
   with a variety of traffic types while measuring the corresponding
   metrics simultaneously becomes necessary.

   Test beds for any type of testing for an NFV-based system will be
   largely similar to previously used test architectures.  The methods
   are impacted by virtualization, as described above, but the design of
   test beds are similar as in the past.  There are two main new

   o  Since networking is based on software, which has lead to greater
      automation in deployment, the test system should also be
      deployable with the rest of the system in order to fully automate
      the system.  This is especially relevant in a DevOps environment
      supported by a Continuous Integration and Continuous Deployment
      (CI/CD) tool chain (see Section 4.11.3 below).

   o  In any performance test bed, the test system should not share the
      same resources as the SUT.  While multi-tenancy is a reality in
      virtualization, having the test system share resources with the
      SUT will impact the measured results in a performance test bed.
      The test system should be deployed on a separate platform in order
      not to impact the resources available to the SUT.

4.11.2.  New Functionality

   NFV presents a collection of new functionality in order to support
   the goal of software networking.  Each component on the architecture
   shown in Figure 2 has an associated set of functionality that allows
   VNFs to run the following: onboarding, life-cycle management for VNFs
   and Network Services (NS), resource allocation, hypervisor functions,

Bernardos, et al.             Informational                    [Page 31]

RFC 8568       Network Virtualization Research Challenges     April 2019

   One of the new capabilities enabled by NFV is VNF Forwarding Graphs
   (VNFFG).  This refers to the graph that represents a network service
   by chaining together VNFs into a forwarding path.  In practice, the
   forwarding path can be implemented in a variety of ways using
   different networking capabilities: vSwitch, SDN, and SDN with a
   northbound application.  Additionally, the VNFFG might use tunneling
   protocols like Virtual eXtensible Local Area Network (VXLAN).  The
   dynamic allocation and implementation of these networking paths will
   have different performance characteristics depending on the methods
   used.  The path implementation mechanism becomes a variable in the
   network testing of the NSs.  The methodology used to test the various
   mechanisms should largely remain the same; as usual, the test
   environment should remain constant for each of the tests, focusing on
   varying the path establishment method.

   "Scaling" refers to the change in allocation of resources to a VNF or
   NS.  It happens dynamically at run-time, based on defined policies
   and triggers.  The triggers can be network, compute, or storage
   based.  Scaling can allocate more resources in times of need, or
   reduce the amount of resources allocated when the demand is reduced.
   The SUT in this case becomes much larger than the VNF itself: MANO
   controls how scaling is done based on policies, and then allocates
   the resources accordingly in the NFVI.  Essentially, the testing of
   scaling includes the entire NFV architecture components into the SUT.

4.11.3.  Opportunities

   Softwarization of networking functionality leads to softwarization of
   the test as well.  As PNFs are being transformed into VNFs, so are
   the test tools.  This leads to the fact that test tools are also
   being controlled and executed in the same environment as the VNFs.
   This presents an opportunity to include VNF-based test tools along
   with the deployment of the VNFs supporting the services of the
   service provider into the host data centers.  Therefore, tests can be
   automatically executed upon deployment in the target environment, for
   each deployment, and each service.  With PNFs, this was very
   difficult to achieve.

   This new concept helps to enable modern concepts like DevOps and
   Continuous Integration and Continuous Deployment in the NFV
   environment.  The CI/CD pipeline supports this concept.  It consists
   of a series of tools, among which immediate testing is an integral
   part, to deliver software from source to deployment.  The ability to
   deploy the test tools themselves into the production environment
   stretches the CI/CD pipeline all the way to production deployment,
   allowing a range of tests to be executed.  The tests can be simple,

Bernardos, et al.             Informational                    [Page 32]

RFC 8568       Network Virtualization Research Challenges     April 2019

   with a goal of verifying the correct deployment and networking
   establishment, but can also be more complex, like testing VNF

5.  Technology Gaps and Potential IETF Efforts

   Table 1 correlates the open network virtualization research areas
   identified in this document to potential IETF and IRTF groups that
   could address some aspects of them.  An example of a specific gap
   that the group could potentially address is identified as a
   parenthetical beside the group name.

   | Open Research Area      | Potential IETF/IRTF Group               |
   | 1) Guaranteeing QoS     | IPPM WG (Measurements of NFVI)          |
   |                         |                                         |
   | 2) Performance          | SFC WG, NFVRG (energy-driven            |
   | improvement             | orchestration)                          |
   |                         |                                         |
   | 3) Multiple Domains     | NFVRG (multi-domain orchestration)      |
   |                         |                                         |
   | 4) Network Slicing      | NVO3 WG, NETSLICES bar BoF (multi-      |
   |                         | tenancy support)                        |
   |                         |                                         |
   | 5) Service Composition  | SFC WG (SFC Mgmt and Config)            |
   |                         |                                         |
   | 6) End-user device      | N/A                                     |
   | virtualization          |                                         |
   |                         |                                         |
   | 7) Security             | N/A                                     |
   |                         |                                         |
   | 8) Separation of        | NFVRG (separation between transport     |
   | control concerns        | control and services)                   |
   |                         |                                         |
   | 9) Testing              | NFVRG (testing of scaling)              |
   |                         |                                         |
   | 10) Function placement  | NFVRG, SFC WG (VNF placement algorithms |
   |                         | and protocols)                          |

     Table 1: Mapping of Open Research Areas to Potential IETF Groups

Bernardos, et al.             Informational                    [Page 33]

RFC 8568       Network Virtualization Research Challenges     April 2019

6.  NFVRG Focus Areas

   Table 2 correlates the currently identified NFVRG topics of interest
   / focus areas to the open network virtualization research areas
   enumerated in this document.  This can help the NFVRG in identifying
   and prioritizing research topics.  The current list of NFVRG focus
   points is the following:

   o  Re-architecting functions, including aspects such as new
      architectural and design patterns (e.g., containerization,
      statelessness, serverless, control/data plane separation), SDN
      integration, and proposals on programmability.

   o  New management frameworks, considering aspects related to new OAM
      mechanisms (e.g., configuration control, hybrid descriptors) and
      lightweight MANO proposals.

   o  Techniques to guarantee low latency, resource isolation, and other
      data-plane features, including hardware acceleration, functional
      offloading to data-plane elements (including NICs), and related

   o  Measurement and benchmarking, addressing both internal
      measurements and external applications.

     | NFVRG Focus Point                   | Open Research Area      |
     | 1) Re-architecting functions        | - Performance improvem. |
     |                                     | - Network Slicing       |
     |                                     | - Guaranteeing QoS      |
     |                                     | - Security              |
     |                                     | - End-user device virt. |
     |                                     | - Separation of control |
     |                                     |                         |
     | 2) New management frameworks        | - Multiple Domains      |
     |                                     | - Service Composition   |
     |                                     | - End-user device virt. |
     |                                     |                         |
     | 3) Low latency, resource isolation, | - Performance improvem. |
     | etc.                                | - Separation of control |
     |                                     |                         |
     | 4) Measurement and benchmarking     | - Guaranteeing QoS      |
     |                                     | - Testing               |

       Table 2: Mapping of NFVRG Focus Points to Open Research Areas

Bernardos, et al.             Informational                    [Page 34]

RFC 8568       Network Virtualization Research Challenges     April 2019

7.  IANA Considerations

   This document has no IANA actions.

8.  Security Considerations

   This is an Informational RFC that details research challenges; it
   does not introduce any security threat.  Research challenges and gaps
   related to security and privacy have been included in Section 4.8.

9.  Informative References

   [COMS-PS]  Geng, L., Slawomir, S., Qiang, L., Matsushima, S., Galis,
              A., and L. Contreras, "Problem Statement of Common
              Operation and Management of Network Slicing", Work in
              Progress, draft-geng-coms-problem-statement-04, March

              Martini, B. and F. Paganelli, "A Service-Oriented Approach
              for Dynamic Chaining of Virtual Network Functions over
              Multi-Provider Software-Defined Networks", Future
              Internet Vol. 8, No. 2, DOI 10.3390/fi8020024, June 2016.

              Clayman, S., Maini, E., Galis, A., Manzalini, A., and
              N. Mazzocca, "The dynamic placement of virtual network
              functions", 2014 IEEE Network Operations and Management
              Symposium (NOMS) pp. 1-9, DOI 10.1109/NOMS.2014.6838412,
              May 2014.

              ETSI NFV ISG, "Network Functions Virtualisation (NFV);
              Terminology for Main Concepts in NFV", ETSI GS NFV 003
              V1.2.1 NFV 003, December 2014,

              ETSI NFV ISG, "Network Functions Virtualisation (NFV);
              Ecosystem; Report on SDN Usage in NFV Architectural
              Framework", ETSI GS NFV-EVE 005 V1.1.1 NFV-EVE 005,
              December 2015,

Bernardos, et al.             Informational                    [Page 35]

RFC 8568       Network Virtualization Research Challenges     April 2019

              ETSI NFV ISG, "Network Functions Virtualisation (NFV); NFV
              Performance & Portability Best Practises", ETSI GS NFV-PER
              001 V1.1.2 NFV-PER 001, December 2014,

              ETSI NFV ISG, "Network Functions Virtualisation (NFV); NFV
              Security; Problem Statement", ETSI GS NFV-SEC 001 V1.1.1
              NFV-SEC 001, October 2014, <http://www.etsi.org/deliver/

              ETSI, "Network Functions Virtualisation (NFV) - White
              Paper #3: Network Operator Perspectives on Industry
              Progress", Issue 1, SDN & OpenFlow World
              Congress Dusseldorf, Germany, October 2014,

              Jain, S., et al., "B4: experience with a globally-deployed
              Software Defined WAN", SIGCOMM '13: Proceedings of the ACM
              SIGCOMM 2013 conference on SIGCOMM, pp. 3-14, Hong
              Kong China, DOI 10.1145/2486001.2486019, August 2013.

              Torre, P., "Discover the Top 10 Differences Between NFV
              and Cloud Environments", November 2015,

              ITU-T, "Y.3300: Framework of software-defined networking",
              ITU-T Recommendation Y.3300, June 2014,

              ITU-T, "Y.3301: Functional requirements of software-
              defined networking", ITU-T Recommendation Y.3301,
              September 2016,

Bernardos, et al.             Informational                    [Page 36]

RFC 8568       Network Virtualization Research Challenges     April 2019

              ITU-T, "Y.3302: Functional architecture of software-
              defined networking", ITU-T Recommendation Y.3302, January
              2017, <http://www.itu.int/rec/T-REC-Y.3302-201701-I/en>.

              Contreras, L., Bernardos, C., Lopez, D., Boucadair, M.,
              and P. Iovanna, "Cooperating Layered Architecture for
              Software Defined Networking (CLAS)", Work in Progress,
              draft-contreras-layered-sdn-03, November 2018.

              Sriram, N., Krishnan, R., Ghanwani, A., Krishnaswamy, D.,
              Willis, P., Chaudhary, A., and F. Huici, "An Analysis of
              Lightweight Virtualization Technologies for NFV", Work in
              Progress, draft-natarajan-nfvrg-containers-for-nfv-03,
              July 2016.

              Bernardos, C., Gero, B., Di Girolamo, M., Kern, A.,
              Martini, B., and I. Vaishnavi, "5GEx: Realizing a Europe-
              wide Multi-domain framework for software-defined
              infrastructures", Transactions on Emerging
              Telecommunications Technologies Vol. 27, No. 9,
              pp. 1271-1280, DOI 10.1002/ett.3085, July 2016.

              Bernardos, C., Contreras, L., Vaishnavi, I., Szabo, R.,
              Li, X., Paolucci, F., Sgambelluri, A., Martini, B.,
              Valcarenghi, L., Landi, G., Andrushko, D., and A. Mourad,
              "Multi-domain Network Virtualization", Work in Progress,
              draft-bernardos-nmrg-multidomain-00, March 2019.

              Galis, A., Dong, J., Makhijani, K., Bryant, S., Boucadair,
              M., and P. Martinez-Julia, "Network Slicing - Introductory
              Document and Revised Problem Statement", Work in
              Progress, draft-gdmb-netslices-intro-and-ps-02, February

   [NFV-COTS] Mo, L. and B. Khasnabish, "NFV Reliability using COTS
              Hardware", Work in Progress, draft-mlk-nfvrg-nfv-
              reliability-using-cots-01, October 2015.

Bernardos, et al.             Informational                    [Page 37]

RFC 8568       Network Virtualization Research Challenges     April 2019

              Luizelli, M., Bays, L., Buriol, L., Barcellos, M., and
              L. Gaspary, "Piecing together the NFV provisioning puzzle:
              Efficient placement and chaining of virtual network
              functions", 2015 IFIP/IEEE International Symposium on
              Integrated Network Management (IM) pp. 98-106,
              DOI 10.1109/INM.2015.7140281, May 2015.

              Mijumbi, R., Serrat, J., Gorricho, J-L., Bouten, N.,
              De Turck, F., and R. Boutaba, "Network Function
              Virtualization: State-of-the-art and Research Challenges",
              IEEE Communications Surveys & Tutorials Volume: 18,
              Issue: 1, pp. 236-262, DOI 10.1109/COMST.2015.2477041,
              September 2015.

              Bagnulo, M. and D. Dolson, "NFVI PoP Network Topology:
              Problem Statement", Work in Progress, draft-bagnulo-nfvrg-
              topology-01, March 2016.

              NGMN Alliance, "NGMN 5G White Paper", Version 1.0,
              February 2015,

   [omniran]  IEEE, "Recommended Practice for Network Reference Model
              and Functional Description of IEEE 802 Access Network",
              P802.1CF IEEE Draft, December 2017.

              Open Networking Foundation, "SDN Architecture", ONF
              TR-521 TR-521, Issue 1.1, February 2016,

   [OpenFlow] Open Networking Foundation, "OpenFlow Switch
              Specification", ONF TS-025, Version 1.5.1 (Protocol
              version 0x06), March 2015.

              Lopez, D., "OpenMANO: The Dataplane Ready Open Source NFV
              MANO Stack", March 2015, <https://www.ietf.org/

Bernardos, et al.             Informational                    [Page 38]

RFC 8568       Network Virtualization Research Challenges     April 2019

   [RFC5810]  Doria, A., Ed., Hadi Salim, J., Ed., Haas, R., Ed.,
              Khosravi, H., Ed., Wang, W., Ed., Dong, L., Gopal, R., and
              J. Halpern, "Forwarding and Control Element Separation
              (ForCES) Protocol Specification", RFC 5810,
              DOI 10.17487/RFC5810, March 2010,

   [RFC6241]  Enns, R., Ed., Bjorklund, M., Ed., Schoenwaelder, J., Ed.,
              and A. Bierman, Ed., "Network Configuration Protocol
              (NETCONF)", RFC 6241, DOI 10.17487/RFC6241, June 2011,

   [RFC7252]  Shelby, Z., Hartke, K., and C. Bormann, "The Constrained
              Application Protocol (CoAP)", RFC 7252,
              DOI 10.17487/RFC7252, June 2014,

   [RFC7426]  Haleplidis, E., Ed., Pentikousis, K., Ed., Denazis, S.,
              Hadi Salim, J., Meyer, D., and O. Koufopavlou, "Software-
              Defined Networking (SDN): Layers and Architecture
              Terminology", RFC 7426, DOI 10.17487/RFC7426, January
              2015, <https://www.rfc-editor.org/info/rfc7426>.

   [RFC7498]  Quinn, P., Ed. and T. Nadeau, Ed., "Problem Statement for
              Service Function Chaining", RFC 7498,
              DOI 10.17487/RFC7498, April 2015,

   [RFC7665]  Halpern, J., Ed. and C. Pignataro, Ed., "Service Function
              Chaining (SFC) Architecture", RFC 7665,
              DOI 10.17487/RFC7665, October 2015,

   [RFC8030]  Thomson, M., Damaggio, E., and B. Raymor, Ed., "Generic
              Event Delivery Using HTTP Push", RFC 8030,
              DOI 10.17487/RFC8030, December 2016,

   [RFC8040]  Bierman, A., Bjorklund, M., and K. Watsen, "RESTCONF
              Protocol", RFC 8040, DOI 10.17487/RFC8040, January 2017,

   [RFC8172]  Morton, A., "Considerations for Benchmarking Virtual
              Network Functions and Their Infrastructure", RFC 8172,
              DOI 10.17487/RFC8172, July 2017,

Bernardos, et al.             Informational                    [Page 39]

RFC 8568       Network Virtualization Research Challenges     April 2019

   [SDN-AAA]  Lopez, R. and G. Lopez-Millan, "Software-Defined
              Networking (SDN)-based AAA Infrastructures Management",
              Work in Progress, draft-marin-sdnrg-sdn-aaa-mng-00,
              November 2015.

              Medhat, A., Taleb, T., Elmangoush, A., Carella, G.,
              Covaci, S., and T. Magedanz, "Service Function Chaining in
              Next Generation Networks: State of the Art and Research
              Challenges", IEEE Communications Magazine vol. 55, no. 2,
              pp. 216-223, DOI 10.1109/MCOM.2016.1600219RP, February

              Foy, X. and A. Rahman, "Network Slicing - 3GPP Use Case",
              Work in Prgoress, draft-defoy-netslices-3gpp-network-
              slicing-02, October 2017.

              Sproule, W. and A. Fernando, "Virtualization of Mobile
              Device User Experience", US Patent 9.542.062 B2, filed
              October 2013 and issued December 2014, Current
              Assignee: Microsoft Technology Licensing LLC.

   [vnf-p]    Moens, H. and , "VNF-P: A model for efficient placement of
              virtualized network functions", 10th International
              Conference on Network and Service Management (CNSM) and
              Workshop pp. 418-423, DOI 10.1109/CNSM.2014.7014205,
              November 2014.

              Rosa, R., Rothenberg, C., and R. Szabo, "VNF Benchmark-as-
              a-Service", Work in Progress, draft-rorosz-nfvrg-vbaas-00,
              October 2015.

              Rosa, R., Rothenberg, C., and R. Szabo, "A VNF Testing
              Framework Design, Implementation and Partial Results",
              NFVRG IETF 97, November 2016,

Bernardos, et al.             Informational                    [Page 40]

RFC 8568       Network Virtualization Research Challenges     April 2019


   The authors want to thank Dirk von Hugo, Rafa Marin, Diego Lopez,
   Ramki Krishnan, Kostas Pentikousis, Rana Pratap Sircar, Alfred
   Morton, Nicolas Kuhn, Saumya Dikshit, Fabio Giust, Evangelos
   Haleplidis, Angeles Vazquez-Castro, Barbara Martini, Jose Saldana,
   and Gino Carrozzo for their very useful reviews and comments to the
   document.  Special thanks to Pedro Martinez-Julia, who provided text
   for the network slicing section.

   The authors want to also thank Dave Oran and Michael Welzl for their
   very detailed IRSG reviews.

   The work of Carlos J. Bernardos and Luis M. Contreras is partially
   supported by the H2020 5GEx (Grant Agreement no. 671636) and
   5G-TRANSFORMER (Grant Agreement no. 761536) projects.

Authors' Addresses

   Carlos J. Bernardos
   Universidad Carlos III de Madrid
   Av. Universidad, 30
   Leganes, Madrid  28911

   Phone: +34 91624 6236
   Email: cjbc@it.uc3m.es
   URI:   http://www.it.uc3m.es/cjbc/

   Akbar Rahman
   InterDigital Communications, LLC
   1000 Sherbrooke Street West, 10th floor
   Montreal, Quebec  H3A 3G4

   Email: Akbar.Rahman@InterDigital.com
   URI:   http://www.InterDigital.com/

   Juan Carlos Zuniga
   425 rue Jean Rostand
   Labege  31670

   Email: j.c.zuniga@ieee.org
   URI:   http://www.sigfox.com/

Bernardos, et al.             Informational                    [Page 41]

RFC 8568       Network Virtualization Research Challenges     April 2019

   Luis M. Contreras
   Telefonica I+D
   Ronda de la Comunicacion, S/N
   Madrid  28050

   Email: luismiguel.contrerasmurillo@telefonica.com

   Pedro Aranda
   Universidad Carlos III de Madrid
   Av. Universidad, 30
   Leganes, Madrid  28911

   Email: pedroandres.aranda@uc3m.es

   Pierre Lynch
   Keysight Technologies
   800 Perimeter Park Dr, Suite A
   Morrisville, NC  27560
   United States of America

   Email: pierre.lynch@keysight.com
   URI:   http://www.keysight.com

Bernardos, et al.             Informational                    [Page 42]