Internet DRAFT - draft-bernini-nfvrg-vnf-orchestration

draft-bernini-nfvrg-vnf-orchestration







NFV Research Group                                            G. Bernini
Internet-Draft                                                  G. Landi
Intended status: Informational                                 Nextworks
Expires: October 12, 2017                                       D. Lopez
                                                              Telefonica
                                                     P. Aranda Gutierrez
                                                                    UC3M
                                                          April 10, 2017


   VNF Pool Orchestration For Automated Resiliency in Service Chains
                draft-bernini-nfvrg-vnf-orchestration-04

Abstract

   Network Function Virtualisation (NFV) aims at evolving the way
   network operators design, deploy and provision their networks by
   leveraging on standard IT virtualisation technologies to move and
   consolidate a wide range of network functions and services onto
   industry standard high volume servers, switches and storage.  The
   primary target for operators, stimulated by the recent updates on NFV
   and SDN, is the network edge.  In fact, operators are considering
   their future datacentres and Points of Presence (PoPs) as
   increasingly dynamic infrastructures where Virtualised Network
   Functions (VNFs) and on-demand chained services with high elasticity
   will be deployed.

   This document presents an orchestration framework for automated
   deployment of highly available VNF chains.  Resiliency of VNFs and
   chained services is a key requirement for operators to improve, ease,
   automate and speed up services lifecycle management.  The proposed
   VNFs orchestration framework is also positioned with respect to
   current NFV and Service Function Chaining (SFC) architectures and
   solutions.

Status of This Memo

   This Internet-Draft is submitted in full conformance with the
   provisions of BCP 78 and BCP 79.

   Internet-Drafts are working documents of the Internet Engineering
   Task Force (IETF).  Note that other groups may also distribute
   working documents as Internet-Drafts.  The list of current Internet-
   Drafts is at http://datatracker.ietf.org/drafts/current/.

   Internet-Drafts are draft documents valid for a maximum of six months
   and may be updated, replaced, or obsoleted by other documents at any




Bernini, et al.         Expires October 12, 2017                [Page 1]

Internet-Draft              VNF Orchestration                 April 2017


   time.  It is inappropriate to use Internet-Drafts as reference
   material or to cite them other than as "work in progress."

   This Internet-Draft will expire on October 12, 2017.

Copyright Notice

   Copyright (c) 2017 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
   (http://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.  Code Components extracted from this document must
   include Simplified BSD License text as described in Section 4.e of
   the Trust Legal Provisions and are provided without warranty as
   described in the Simplified BSD License.

Table of Contents

   1.  Introduction  . . . . . . . . . . . . . . . . . . . . . . . .   2
   2.  Terminology . . . . . . . . . . . . . . . . . . . . . . . . .   4
   3.  VNF Pool Orchestration for Resilient Virtual Appliances . . .   4
     3.1.  Problem Statement . . . . . . . . . . . . . . . . . . . .   5
     3.2.  Orchestration Framework . . . . . . . . . . . . . . . . .   6
       3.2.1.  Orchestrator  . . . . . . . . . . . . . . . . . . . .   8
       3.2.2.  SDN Controller  . . . . . . . . . . . . . . . . . . .   8
       3.2.3.  Service Function Path Manager . . . . . . . . . . . .   9
       3.2.4.  Edge Configurator . . . . . . . . . . . . . . . . . .  10
     3.3.  Resiliency Control Functions for Chained VNFs . . . . . .  10
   4.  Positioning in Existing NFV and SFC Frameworks  . . . . . . .  12
     4.1.  Mapping into NFV Architecture . . . . . . . . . . . . . .  12
     4.2.  Mapping into SFC Architecture . . . . . . . . . . . . . .  13
   5.  IANA Considerations . . . . . . . . . . . . . . . . . . . . .  13
   6.  Security Considerations . . . . . . . . . . . . . . . . . . .  13
   7.  Acknowledgements  . . . . . . . . . . . . . . . . . . . . . .  13
   8.  References  . . . . . . . . . . . . . . . . . . . . . . . . .  14
     8.1.  Normative References  . . . . . . . . . . . . . . . . . .  14
     8.2.  Informative References  . . . . . . . . . . . . . . . . .  14
   Authors' Addresses  . . . . . . . . . . . . . . . . . . . . . . .  14

1.  Introduction

   Current Telco infrastructures are facing the rapid development of the
   cloud market, which includes a broad range of emerging virtualised
   services and distributed applications.  Network Function



Bernini, et al.         Expires October 12, 2017                [Page 2]

Internet-Draft              VNF Orchestration                 April 2017


   Virtualisation (NFV) is gaining wide interest across operators as a
   means to evolve the way networks are operated and provisioned, with
   network functions and services traditionally integrated in hardware
   devices executed in virtualised environments.

   A Virtualised Network Function (VNF) provides the same function as
   its non virtualised equivalent (e.g. firewall, load balancer) but is
   deployed as a software instance running on general purpose servers
   using virtualisation technologies.  The main idea, therefore, is to
   run network functions in datacentres or commodity network nodes that
   are, in some cases, close to the end user premises.  With NFV,
   network functions are moved from specialised hardware devices to
   self-contained virtual machines running in general purpose servers.
   These virtualised functions can be deployed in multiple instances or
   moved to various locations in the network, adapting themselves to
   traffic dynamicity and customer demands without the overhead cost and
   management of installing new equipment.

   Operator networks are populated with a large and increasing variety
   of proprietary software and hardware tools and appliances.  The
   deployment of new network services in operational environments is
   often a complex and costly procedure, where additional physical space
   and power are required to accommodate new boxes.  Additionally,
   current hardware-based appliances rapidly reach end of life.  This
   requires that much of the design integration and deployment cycle be
   repeated with little revenue benefit.  In this context, the
   transition of network functions and appliances from hardware to
   software solutions by means of NFV promises to address and overcome
   these hindrances for network operators.

   The considerations above are valid for stand-alone VNFs running
   independently.  However, additional challenges and requirements raise
   for network operators when services offered to customers are built by
   the composition of multiple VNFs.  In this case, the deployment and
   provisioning of each (virtual) service component for the customer
   needs to be coordinated with the other VNFs, applying control
   functions to steer the traffic through them following a predefined
   order (i.e. according to the specific service function path).  An
   orchestration framework capable of coordinating the automated
   deployment, configuration, provisioning and chaining of multiple VNFs
   would ease the management of the whole lifecycle of services offered
   to customers.  Additionally, when dealing with virtualised functions,
   resiliency and high availability of chained services pose additional
   requirements for a VNF orchestration framework, in terms of detection
   of software failures at various levels (including hypervisors and
   virtual machines, hardware failure), and dynamic and intellingent
   reaction (virtual appliance migration, deployment of new VNFs, re-
   adapt the VNF chain).



Bernini, et al.         Expires October 12, 2017                [Page 3]

Internet-Draft              VNF Orchestration                 April 2017


   This document presents an orchestration framework for automated
   deployment of high available VNF chains, and introduces its
   architecture and building blocks.  Resiliency for both stand-alone
   VNFs and chained services is considered in this document as a key
   control function based on VNF pool concepts.  The proposed VNF pool
   orchestration framework is also positioned with respect to approaches
   and architectures currently defined for Network Function
   Virtualisation and Service Function Chaining (SFC).

2.  Terminology

   The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
   "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
   document are to be interpreted as described in [RFC2119].

   The following acronyms are used in this document:

   NFV:  Network Function Virtualisation.

   SDN:  Software Defined Networking.

   VNF:  Virtualised Network Function.

   SFC:  Service Function Chaining.

   CPE:  Customer Premise Equipment.

   VPN:  Virtual Private Network.

   EMS:  Element Management System.

   PoP:  Point of Presence.

   VM:  Virtual Machine.

3.  VNF Pool Orchestration for Resilient Virtual Appliances

   The telco market is rapidly moving towards an "Everything as a
   Service" model, where the virtualisation of traditionally in-the-box
   network functions can benefit from Software Defined Networking (SDN)
   tools and technologies.  As said, the recent updates and proposed
   solutions on NFV and SDN is practically bringing, from an operator
   perspective, a deep evolution on how the network edge is architected,
   operated and provisioned, since that is the place where VNFs and
   virtual services can be deployed and provisioned close to the
   customers.  Operators target to evolve their datacentres and PoPs
   into increasingly dynamic infrastructures where VNFs and chained
   services can be deployed with high availability and high elasticity



Bernini, et al.         Expires October 12, 2017                [Page 4]

Internet-Draft              VNF Orchestration                 April 2017


   to scale up and down while optimizing performances and resources
   utilization.

   This section introduces the VNF pool orchestration framework for the
   deployment, provisioning and chaining of resilient virtual appliances
   and services within operator data-centres proposed in this document.

3.1.  Problem Statement

   The orchestration framework proposed in this document aims at solving
   some of the challenges that operators face when, trying to apply the
   base NFV concepts, they replace hardware devices implementing well-
   known network functions with software-based virtual appliances.  In
   particular, this VNF orchestration framework targets an automated,
   flexible and elastic provisioning of service chains within operators'
   datacentres.

   When operators need to compose and chain multiple VNFs to provision a
   given service to the customer, they need to operate network and
   computing resources in a coordinated way, and above all to implement
   control mechanisms and procedures to steer the traffic through the
   different VNFs and the customer sites.  As an example, the
   virtualisation of the Customer Premises Equipment (CPE) is emerging
   as one of the first applications of the Network Functions
   Virtualisation (NFV) architecture that is currently being
   commercialized by several software (and hardware) vendors.  It has
   the potential to generate a significant impact on the operators
   businesses.  The term virtual CPE (vCPE) refers to the execution in a
   virtual environment of the network functions that traditionally
   integrated in hardware gears at customer premises, like BGP speakers,
   firewall, NAT, etc.

   Different scenarios and use cases exist for the vCPE.  Currently, the
   typical scenario is the vCPE in the PoP, that actually provides
   softwarization and shift to the first PoP of the operator for those
   network functions normally deployed at customer premises (e.g.  NAT,
   Firewall, etc.).  The goal is to manage the whole LAN environment of
   the customer, while preserving QoE of its services, providing added
   value services to users not willing to get involved with technology
   issues, and reducing maintenance trouble tickets and the need for in-
   house problem solving.  In addition, the vCPE can be used in the
   operator's datacenter to implement in software those chained network
   functions to provision automated VPN services for customers
   [VCPE-T2], in order to dynamically and automatically extend existing
   L3 VPNs (e.g. connecting remote customer sites) to incorporate new
   virtual assets (like virtual machines) into a private cloud.





Bernini, et al.         Expires October 12, 2017                [Page 5]

Internet-Draft              VNF Orchestration                 April 2017


   As additional requirements for the proposed orchestration framework,
   the use of VNFs opens new challenges concerning the reliability of
   provided virtual services.  When network functions are deployed on
   monolithic hardware platforms, the lifecycle of individual services
   is strictly bound to the availability of the physical device, and
   management tools may detect outages and migrate affected services to
   new instances deployed on backup hardware.  When introducing VNFs,
   individual network functions may still fail, but with more risk
   factors such as software failure at various levels, including
   hypervisors and virtual machines, hardware failure, and virtual
   appliance migration.  Moreover, when considering chains of VNFs, the
   management and control tools used by the operators have to consider
   and apply reliability mechanisms at the service level, including
   transparent migration to backup VNFs and synchronization of state
   information.  In this context, VNF pooling mechanisms and concepts
   are valid and applicable, thus considering VNF instances grouped as
   pools to provide the same function in a reliable way.

3.2.  Orchestration Framework

   The VNF pool orchestration framework proposed in this document aims
   to provide automated functions for the deployment, provisioning and
   composition of resilient VNFs within operators' datacentres.
   Figure Figure 1 presents the high level architecture, including
   building blocks and functional components.

   This VNF pool orchestration framework is built around two key
   components: the orchestrator and the SDN controller.  The
   orchestrator includes all the functions related to the management,
   coordination, and control of VNFs instantiation, configuration and
   composition.  It is the component at the highest level of the
   architecture and represents the access point to the VNF pool
   orchestration framework for the operator.  On the other hand, the SDN
   controller provides dynamic traffic steering and flexible network
   provisioning within the datacenter as needed by the VNF chains.  The
   basic controller functions are augmented by a set of enhanced network
   applications deployed on top, that might be themselves control and
   management VNFs for operator use (i.e. not related to customers and
   users functions).

   Therefore, the architecture depicted in Figure Figure 1 is a
   practical demonstration of how SDN and NFV technologies and concepts
   can be integrated to provide substantial benefits to network
   operators in terms of robustness, ease of management, control and
   provisioning of their network infrastructures and services.  SDN and
   NFV are clearly complementary solutions for enabling virtualisation
   of network infrastructures, services and functions while supporting
   dynamic and flexible network traffic engineering.



Bernini, et al.         Expires October 12, 2017                [Page 6]

Internet-Draft              VNF Orchestration                 April 2017


                     +----------------------------------------------+
                     |                Orchestrator                  |
                     +-----+----------------+----------------+------+
                           |                |                |
                           V                V                V
                     +------------+  +-------------+  +-------------+
                     |            |  |             |  |             |
                     |  VNF Pool  |  | Service Fcn |  |    Edge     |
                     |  Manager   |  |  Path Mgr   |  |Configurator |
                     |            |  |             |  |             |
                     +-----+------+  +------+------+  +------+------+
                           |                |                |
                           V                V                V
                     +----------------------------------------------+
                     |                 SDN Controller               |
                     +--------------+---+----+---------------+------+
+.......................+           |   |    |              |
:           +--------+  \           |   |    |              +-----+
:           |+-------++ :|    +-----+---+----+----+               |
:           +|       || :\    | +-----------------+-+         +---+----+
:          /|| VNF1  || : |   | | +---------------+-+-+       |        |
:         / ||       || :..+  | | |    Virtual    | | +-------+  Edge  |
:        /  ++-------+| :  :  | | |   Switch(es)  | | +-------+ Router |
:       /    +---+----+ :  :| +-+-+---------------+ | |       |        |
:      /         |      :  :\   +-+-----------------+ |       +--------+
:  +--+-----+    |      :  : |    +-----+---+---+-----+
:  |+-------++   |      :  : \          |   |   |
:  ||       ||   |      :  :  |         |   |   |
:  || VNF2  ||   |      :  :  \---------+---+---+------+
:  ||       ||   |      :  :  |                        |
:  ++-------+|   |      :  :  |                        |
:   +--+-+---+   |      :  :  |     Virtual Compute    |
:         \ +----+---+  :  :  |     Infrastructure     |
:          \|+-------++ :  :  |                        |
:           +|       || :  :  |                        |
:           || VNF3  || :  :  /------------------------+
:           ||       || :  : |
:           ++-------+| :  : |
:            +--------+ :  : |
:Service Chain "X"      :  : /
+.......................+  : |
  :Service Chain "Y"       :/
  +........................+


         Figure 1: VNF Pool Orchestration Framework Architecture.





Bernini, et al.         Expires October 12, 2017                [Page 7]

Internet-Draft              VNF Orchestration                 April 2017


   SDN focuses on network programmability, traffic steering and multi-
   tenancy by means of a common, open, and dynamic abstraction of
   network resources.  NFV targets a progressive migration of network
   elements, network appliances and fixed function boxes into VMs that
   can be ran on commodity hardware, enabling the benefits of cloud and
   datacentres to be applied to network functions.

3.2.1.  Orchestrator

   The VNF orchestrator implements a set of functions to seamlessly
   control and manage in a coordinated way the instantiation and
   deployment of VNFs on one hand, and their composition and chaining to
   steer the traffic through them on the other.  It is fully controlled
   and operated by the network operator, and basically it is the highest
   control and orchestration layer that sits above all the softwarized
   and virtualised components in the proposed architecture.

   Therefore the VNF orchestrator provides a consistent way to access
   the system and provision chains of VNFs to the operator.  It exposes
   a set of primitives to instantiate, configure VNFs and compose them
   according to the specific service chain requirements.  Practically,
   it aims at enabling an efficient and dynamic management of operator's
   infrastructure resources with great flexibility by means of a
   consistent set of APIs.

   To enable this, the VNF orchestrator can be seen as a composition of
   several internal functionalities, each providing a given coordination
   function needed to orchestrate the lower layer control and management
   functions depicted in Figure Figure 1 (i.e.  VNF chain configuration,
   VNF pool provisioning, etc.).  In practice, the VNF orchestrator
   needs to include at least an internal component to manage the
   instantiation and configuration of stand-alone VNFs (e.g. implemented
   by a self-contained VM) that might be directly interfaced with the
   physical servers in the datacenter.  And also a dedicated component
   for programmatic coordination and provisioning of VNF chains is
   needed to properly orchestrate the traffic steering through VNFs
   belonging to the same service chain.  This should also provide multi-
   tenant functionalities and maintain isolation across VNF chains
   deployed for different customers.  It is then clear that the VNF
   orchestrator is the overall coordinator of the proposed framework,
   and it drives all the lower layer components that implement the
   actual control logic.

3.2.2.  SDN Controller

   The SDN controller provides the logic for network control,
   provisioning and monitoring.  It is the component where the SDN
   abstraction happens.  This means it exposes a set of primitives to



Bernini, et al.         Expires October 12, 2017                [Page 8]

Internet-Draft              VNF Orchestration                 April 2017


   configure the datacenter network according to the requirements of the
   VNF chains to be provisioned, while hiding the specific technology
   constraints and capabilities of the software switches and edge
   routers underneath.  The deployment of an SDN controller allows to
   implement a software driven VNF orchestration, with flexible and
   programmable network functions for service chaining and resilient
   virtual appliances.

   At its southbound interface, the SDN controller interfaces with
   software switches running in servers, physical switches
   interconnecting them and edge routers connecting the datacenter with
   external networks.  Multiple control protocols can be used at this
   southbound interface to actually provision the datacenter network and
   enable traffic steering through VNFs, including OpenFlow, OVSDB,
   NETCONF and others.

   Therefore the SDN controller provides the basic network provisioning
   functions needed by upper layer coordination functions to perform
   service chain and VNF pool-wide actions.  Indeed, the logic and the
   state at the service level is only maintained and coordinated by
   network applications on top of the SDN controller.

3.2.3.  Service Function Path Manager

   The service function path manager is deployed as a bridging component
   between the orchestrator and the SDN controller, and it is mostly
   dedicated to the implementation of VNF chaining and composition
   logic.  It computes a suitable path to interconnect the involved VNFs
   (already instantiated and identified by the orchestrator) and
   forwards the network configuration request to the SDN controller for
   each new VNF chain requested by the orchestrator.

   Following the datacenter service chains and related traffic types
   defined in [I-D.ietf-sfc-dc-use-cases], the service function path
   manager should implement its coordination logic to support both
   north-south and east-west chains.  The former refer to network
   traffic staying within the datacenter but coming from a remote
   datacenter or a user through the edge router connecting to an
   external network.  In this case, the service function path manager
   should also coordinate with the edge configurator to properly
   provision the datacenter edge router.  Moreover, this north-south
   case may also refer to VNF chains spanning multiple datacentres, thus
   requiring a further inter-datacenter coordination between service
   function path managers and orchestrators.  These coordination
   functions are out of the scope of this document.  On the other hand,
   the east-west chains refer to VNFs treating network traffic that do
   not exit the datacenter.  For both cases, the service function path
   manager (in combination with the SDN controller) should implement and



Bernini, et al.         Expires October 12, 2017                [Page 9]

Internet-Draft              VNF Orchestration                 April 2017


   support proper service chains encapsulation solutions
   [I-D.ietf-sfc-nsh] to isolate and segregate traffic related to VNF
   chains belonging to different tenants.

   Different deployment models may exist for the service function path
   manager: a dedicated configurator for each chain, or a single
   configurator for all the VNF chains.  In the first approach, the
   orchestrator needs to implement some coordination logic related to
   the dynamic instantiation of configurators when new VNF chains are
   provisioned.

3.2.4.  Edge Configurator

   The edge configurator is a network control application deployed on
   top of the SDN controller.  Its main role is to coordinate the
   provisioning and configuration of the edge router for those north-
   south VNF chains exiting the datacenter.  In particular, it keeps the
   binding between the traffic steered through the VNFs and the related
   network service outside the datacenter terminated at the edge router
   (e.g. a L3 VPN, VLAN, VXLAN, VRF etc), possibly considering the
   service chain encapsulation implemented within the VNF chain.  The
   mediation of the SDN controller allows to support a variety of
   control and management protocols for the actual configuration of the
   datacenter edge router.

3.3.  Resiliency Control Functions for Chained VNFs

   In the proposed orchestration architecture, the resiliency control
   functions that have been identified as a key feature for a flexible
   and dynamic provisioning of chained VNF services are implemented by
   the VNF pool manager depicted in Figure Figure 1.  It is the entity
   that manages and coordinates VNFs reliability providing high
   availability and resiliency features at both stand-alone and chained
   VNFs level.

   The deployment of VNF based services requires moving the resiliency
   capabilities and mechanisms from physical network devices (which are
   typically highly available and often specialized) to entities (like
   self-contained VMs) running VNFs in the context of pools of
   virtualised resources.  When moving towards a resilient approach for
   VNF deployment and operation, in line with ETSI NFV Resiliency
   Requirements (NFV-REL001), the generic high availability requirements
   to be matched are translated into:

   Service continuity:  when a hardware failure or capacity limits
      (memory and CPU) occur on platforms hosting VMs (and therefore
      VNFs), it is necessary to migrate VNFs to other VMs and/or




Bernini, et al.         Expires October 12, 2017               [Page 10]

Internet-Draft              VNF Orchestration                 April 2017


      hardware platforms to guarantee service continuity with minimum
      impact on the users

   Topological transparency:  the hand-over between live and backup VNFs
      must be implemented in a transparent way for the user and also for
      the service chain itself.  The backup VNF instances need to
      replicate the necessary information (configuration, addressing,
      etc.) so that the network function is taken over without any
      topological disruption (i.e. at the VNF chain level)

   Load balancing or scaling:  migration of VNF instances may also
      happen for load-balancing purposes (e.g. for CPU, memory overload
      in virtualised platforms) or scaling of network services (with
      VNFs moved to new hardware platforms).  In both cases the working
      network function is moved to a new VNF instance and the service
      continuity must be maintained.

   Auto scale of VNFs instances:  when a VNF requires increased resource
      allocation to improve overall service performance, the network
      function could be distributed across multiple VMs, and to
      guarantee the performance improvement dedicated pooling mechanisms
      for scaling up or down resources to each VNF in a consistent way
      are needed.

   Multiple VNF resiliency classes:  each type of end-to-end service
      (e.g. web, financial backend, video streaming, etc.) has its own
      specific resiliency requirements for the related VNFs.  While for
      operators it is not easy to achieve service resiliency SLAs
      without building to peak, a basic set of VNF resiliency classes
      can be defined to identify some metrics, such as: if a VNF needs
      status synchronization; fault detection and restoration time
      objective (e.g. real-time); service availability metrics; service
      quality metrics; service latency metrics for VNF chain components.

   The aim of the VNF pool orchestration presented in this document is
   to address the above requirements by introducing the VNF pool manager
   that follows the principles of the IETF VNFPOOL architecture [I-
   D.zong-vnfpool-arch], where a pool manager coordinates the
   reliability of stand-alone VNFs, by selecting the active instance and
   interacting with the Service Control Entity for consistent end-to-end
   service chain reliability and provisioning.  In the VNF pool
   orchestration architecture illustrated in Figure Figure 1, the
   Service Control Entity is implemented by the combination of the
   orchestrator (for overall coordination of service chains) and the VNF
   chain configuration (for actual provisioning and coordination of
   individual service chains).





Bernini, et al.         Expires October 12, 2017               [Page 11]

Internet-Draft              VNF Orchestration                 April 2017


   Different deployment models may exist for the VNF pool manager: a
   dedicated manager for each VNF chain, or a single one for all the
   chains.

   In terms of offered resiliency functionalities, the VNF pool manager
   provides some post-configuration functions to instantiate VNFs (as
   self-contained VMs) with the desired degree of reliability and
   redundancy.  This translates into further actions to create and
   configure additional VMs as backups, therefore building a pool for
   each VNF in the chain.

   The VNF pool manager is conceived to offer several types and degrees
   of reliability functions.  First, it provides specific functions for
   the persistence of VNFs configuration, including making periodic
   snapshots of the VMs running the VNF.  Moreover, at runtime (i.e.
   with the service chain in place), it monitors the operational status
   and performances of the master VNFs VMs, and collects notifications
   about VMs status, e.g. by registering as an observer to dedicated
   services offered by the virtualisation platform used within the
   virtual compute infrastructure.  Moreover, VNF pool manager reacts to
   any failure condition by autonomously replacing the master VNF with
   one of its backup on the pool, basically implementing a swap of VMs
   for service chain recovery purposes.  Thus, the VNF pool manager also
   takes care in coordination with the service function path manager of
   implementing those resiliency mechanisms at the chain level.  Two
   options have been identified so far: cold recovery and hot recovery.
   In the former, backup VNFs, properly configured with the same master
   configuration, are kept ready (but switched off) to be started when
   the master dies.  In this case the recovery time depends on the
   specific VNF and its type of function, e.g. it may depend on
   convergence time for a virtual BGP router.  In the hot recovery,
   active backup VNFs are kept synchronized with the master ones, and
   the recovery of the service chain (mostly performed at the service
   function path manager) in case of failure is faster than cold
   recovery.

4.  Positioning in Existing NFV and SFC Frameworks

4.1.  Mapping into NFV Architecture

   For the presented solution to be integrated in the ETSI NFV reference
   architecture, some modifications need to be applied to it, with main
   focus on the Management and Orchestration (MANO) functions.  The VNF
   pools replace the VNFs in the architecture.  They are then controlled
   by the Element Management System (EMS) on the northbound.  So the EMS
   has to be made VNFPOOL aware.  Additional elements that need to
   support the mechanisms proposed by VNFPOOL are the VNF managers,
   which need to implement the resiliency and VNF scaling



Bernini, et al.         Expires October 12, 2017               [Page 12]

Internet-Draft              VNF Orchestration                 April 2017


   (up-/downscale) functions.  This has also implications on the NFV
   Orchestrator, which has to be aware of the augmented functionality
   offered by the VNF Manager.  In fact, the NFV Orchestrator also
   provides primitives for VNFs chaining, matching the Service Control
   Entity in the VNFPool architecture.  Therefore, even if ETSI NFV MANO
   does not include explicitely SDN, the Edge Configurator, and part of
   the Service Function Path Manager features might be also covered by
   an augmented NFV Orchestrator.

4.2.  Mapping into SFC Architecture

   [I-D.ietf-sfc-architecture] describes the Service Function Chaining
   (SFC) architecture.  It describes the concept of a service function
   (SF) and how to chain SFs and provides only little detail of the SFC
   control plane, which is responsible with the coordination of the SFs
   and their stitching into SFCs.  The combination of orchestrator,
   service function path manager and VNF pool manager functionalities
   described in this document cover most of the functions expected from
   the SFC control plane.

   The interaction with the SFC Classifier is left for further study.
   We expect the VNFPOOL architecture to leverage on it to make sure
   that all VNFPOOL instances will be served traffic on during scale-up
   and that no traffic will be lost during scale-down.

5.  IANA Considerations

   This draft does not have any IANA consideration.

6.  Security Considerations

   Security issues related to VNF pool orchestration and resiliency of
   service chains are left for further study.

7.  Acknowledgements

   This work has been partially supported by the European Commission
   through the H2020 5G Crosshaul (The integrated fronthaul/backhaul,
   grant agreement no:H2020-671598) and Selfnet (Framework for Self-
   organized network management in virtualized and software defined
   networks, grant agreement no:H2020-671672) projects.  The views
   expressed here are those of the authors only.  The European
   Commission is not liable for any use that may be made of the
   information in this document.

   Authors would also like to thank V.  Maffione and G.  Carrozzo from
   Nextworks for valuable discussions and contributions to the topics
   addressed in this document.



Bernini, et al.         Expires October 12, 2017               [Page 13]

Internet-Draft              VNF Orchestration                 April 2017


8.  References

8.1.  Normative References

   [RFC2119]  Bradner, S., "Key words for use in RFCs to Indicate
              Requirement Levels", BCP 14, RFC 2119,
              DOI 10.17487/RFC2119, March 1997,
              <http://www.rfc-editor.org/info/rfc2119>.

8.2.  Informative References

   [I-D.ietf-sfc-dc-use-cases]
              Surendra, S., Tufail, M., Majee, S., Captari, C., and S.
              Homma, "Service Function Chaining Use Cases In Data
              Centers", draft-ietf-sfc-dc-use-cases-06 (work in
              progress), February 2017.

   [I-D.ietf-sfc-architecture]
              Halpern, J. and C. Pignataro, "Service Function Chaining
              (SFC) Architecture", draft-ietf-sfc-architecture-11 (work
              in progress), July 2015.

   [I-D.ietf-sfc-nsh]
              Quinn, P. and U. Elzur, "Network Service Header", draft-
              ietf-sfc-nsh-12 (work in progress), February 2017.

   [I-D.zong-vnfpool-problem-statement]
              Zong, N., Dunbar, L., Shore, M., Lopez, D., and G.
              Karagiannis, "Virtualized Network Function (VNF) Pool
              Problem Statement", draft-zong-vnfpool-problem-
              statement-06 (work in progress), July 2014.

   [VCPE-T2]  G. Bernini, G. Carrozzo, P. A. Gutierrez, D. R. Lopez, ,
              "Virtualising the Network Edge: Virtual CPE for the
              datacenter and the PoP", European Conference on Networks
              and Communications , June 2014.

Authors' Addresses

   Giacomo Bernini
   Nextworks
   Via Livornese 1027
   San Piero a Grado, Pisa  56122
   Italy

   Phone: +39 050 3871600
   Email: g.bernini@nextworks.it




Bernini, et al.         Expires October 12, 2017               [Page 14]

Internet-Draft              VNF Orchestration                 April 2017


   Giada Landi
   Nextworks
   Via Livornese 1027
   San Piero a Grado, Pisa  56122
   Italy

   Phone: +39 050 3871600
   Email: g.landi@nextworks.it


   Diego R. Lopez
   Telefonica
   C. Zurbaran, 12
   Madrid  28010
   Spain

   Email: diego.r.lopez@telefonica.com


   Pedro A. Aranda Gutierrez
   Universidad Carlos III Madrid
   Leganes  28911
   Spain

   Email: paranda@it.uc3m.es


























Bernini, et al.         Expires October 12, 2017               [Page 15]