RFC 7665






Internet Engineering Task Force (IETF)                   J. Halpern, Ed.
Request for Comments: 7665                                      Ericsson
Category: Informational                                C. Pignataro, Ed.
ISSN: 2070-1721                                                    Cisco
                                                            October 2015


              Service Function Chaining (SFC) Architecture

Abstract



   This document describes an architecture for the specification,
   creation, and ongoing maintenance of Service Function Chains (SFCs)
   in a network.  It includes architectural concepts, principles, and
   components used in the construction of composite services through
   deployment of SFCs, with a focus on those to be standardized in the
   IETF.  This document does not propose solutions, protocols, or
   extensions to existing protocols.

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 Engineering Task Force
   (IETF).  It represents the consensus of the IETF community.  It has
   received public review and has been approved for publication by the
   Internet Engineering Steering Group (IESG).  Not all documents
   approved by the IESG are a candidate for any level of Internet
   Standard; see Section 2 of RFC 5741.

   Information about the current status of this document, any errata,
   and how to provide feedback on it may be obtained at
   http://www.rfc-editor.org/info/rfc7665.

















Halpern & Pignataro           Informational                     [Page 1]

RFC 7665                    SFC Architecture                October 2015


Copyright Notice



   Copyright (c) 2015 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.





































Halpern & Pignataro           Informational                     [Page 2]

RFC 7665                    SFC Architecture                October 2015


Table of Contents



   1.  Introduction  . . . . . . . . . . . . . . . . . . . . . . . .   4
     1.1.  Scope . . . . . . . . . . . . . . . . . . . . . . . . . .   4
     1.2.  Assumptions . . . . . . . . . . . . . . . . . . . . . . .   5
     1.3.  Specification of Requirements . . . . . . . . . . . . . .   5
     1.4.  Definition of Terms . . . . . . . . . . . . . . . . . . .   6
   2.  Architectural Concepts  . . . . . . . . . . . . . . . . . . .   8
     2.1.  Service Function Chains . . . . . . . . . . . . . . . . .   8
     2.2.  Service Function Chain Symmetry . . . . . . . . . . . . .   9
     2.3.  Service Function Paths  . . . . . . . . . . . . . . . . .  10
       2.3.1.  Service Function Chains, Service Function Paths, and
               Rendered Service Path . . . . . . . . . . . . . . . .  11
   3.  Architecture Principles . . . . . . . . . . . . . . . . . . .  12
   4.  Core SFC Architecture Components  . . . . . . . . . . . . . .  13
     4.1.  SFC Encapsulation . . . . . . . . . . . . . . . . . . . .  14
     4.2.  Service Function (SF) . . . . . . . . . . . . . . . . . .  15
     4.3.  Service Function Forwarder (SFF)  . . . . . . . . . . . .  15
       4.3.1.  Transport-Derived SFF . . . . . . . . . . . . . . . .  17
     4.4.  SFC-Enabled Domain  . . . . . . . . . . . . . . . . . . .  17
     4.5.  Network Overlay and Network Components  . . . . . . . . .  18
     4.6.  SFC Proxy . . . . . . . . . . . . . . . . . . . . . . . .  18
     4.7.  Classification  . . . . . . . . . . . . . . . . . . . . .  19
     4.8.  Reclassification and Branching  . . . . . . . . . . . . .  19
     4.9.  Shared Metadata . . . . . . . . . . . . . . . . . . . . .  20
   5.  Additional Architectural Concepts . . . . . . . . . . . . . .  21
     5.1.  The Role of Policy  . . . . . . . . . . . . . . . . . . .  21
     5.2.  SFC Control Plane . . . . . . . . . . . . . . . . . . . .  21
     5.3.  Resource Control  . . . . . . . . . . . . . . . . . . . .  22
     5.4.  Infinite Loop Detection and Avoidance . . . . . . . . . .  23
     5.5.  Load-Balancing Considerations . . . . . . . . . . . . . .  23
     5.6.  MTU and Fragmentation Considerations  . . . . . . . . . .  24
     5.7.  SFC OAM . . . . . . . . . . . . . . . . . . . . . . . . .  25
     5.8.  Resilience and Redundancy . . . . . . . . . . . . . . . .  26
   6.  Security Considerations . . . . . . . . . . . . . . . . . . .  27
   7.  References  . . . . . . . . . . . . . . . . . . . . . . . . .  29
     7.1.  Normative References  . . . . . . . . . . . . . . . . . .  29
     7.2.  Informative References  . . . . . . . . . . . . . . . . .  29
   Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . .  30
   Contributors  . . . . . . . . . . . . . . . . . . . . . . . . . .  31
   Authors' Addresses  . . . . . . . . . . . . . . . . . . . . . . .  32










Halpern & Pignataro           Informational                     [Page 3]

RFC 7665                    SFC Architecture                October 2015


1.  Introduction



   The delivery of end-to-end services often requires various service
   functions.  These include traditional network service functions such
   as firewalls and traditional IP Network Address Translators (NATs),
   as well as application-specific functions.  The definition and
   instantiation of an ordered set of service functions and subsequent
   "steering" of traffic through them is termed Service Function
   Chaining (SFC).

   This document describes an architecture used for the creation and
   ongoing maintenance of Service Function Chains (SFCs) in a network.
   It includes architectural concepts, principles, and components, with
   a focus on those to be standardized in the IETF.  SFCs enable
   composite services that are constructed from one or more service
   functions.

   An overview of the issues associated with the deployment of end-to-
   end service function chains, abstract sets of service functions and
   their ordering constraints that create a composite service, and the
   subsequent "steering" of traffic flows through said service
   functions, is described in [RFC7498].

   The current service function deployment models are relatively static,
   coupled to network topology and physical resources, greatly reducing
   or eliminating the ability of an operator to introduce new services
   or dynamically create service function chains.  This architecture
   presents a model addressing the problematic aspects of existing
   service deployments, including topological independence and
   configuration complexity.

1.1.  Scope



   This document defines the architecture for Service Function Chaining
   (SFC) as standardized in the IETF.  The SFC architecture is
   predicated on topological independence from the underlying forwarding
   topology.

   In this architecture, packets are classified on ingress for handling
   by the required set of Service Functions (SFs) in the SFC-enabled
   domain and are then forwarded through that set of functions for
   processing by each function in turn.  Packets may be reclassified as
   a result of this processing.

   The architecture described in this document is independent of the
   planned usage of the network and deployment context and thus, for
   example, is applicable to both fixed and mobile networks as well as
   being useful in many data center applications.



Halpern & Pignataro           Informational                     [Page 4]

RFC 7665                    SFC Architecture                October 2015


   The architecture described herein is assumed to be applicable to a
   single network administrative domain.  While it is possible for the
   architectural principles and components to be applied to inter-domain
   SFCs, these are left for future study.

1.2.  Assumptions



   The following assumptions are made:

   o  There is no standard definition or characterization applicable to
      all SFs, and thus the architecture considers each SF as an opaque
      processing element.

   o  There is no global or standard list of SFs enabled in a given
      administrative domain.  The set of SFs enabled in a given domain
      is a function of the currently active services that may vary with
      time and according to the networking environment.

   o  There is no global or standard SF chaining logic.  The ordered set
      of SFs that needs to be applied to deliver a given service is
      specific to each administrative entity.

   o  The chaining of SFs and the criteria to invoke them are specific
      to each administrative entity that operates an SF-enabled domain.

   o  Several SF chaining policies can be simultaneously applied within
      an administrative domain to meet various business requirements.

   o  The underlay is assumed to provide the necessary connectivity to
      interconnect the Service Function Forwarders (SFFs; see
      Section 1.4), but the architecture places no constraints on how
      that connectivity is realized other than it have the required
      bandwidth, latency, and jitter to support the SFC.

   o  No assumption is made on how Forwarding Information Bases (FIBs)
      and Routing Information Bases (RIBs) of involved nodes are
      populated.

   o  How to bind traffic to a given SF chain is policy-based.

1.3.  Specification of Requirements



   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].






Halpern & Pignataro           Informational                     [Page 5]

RFC 7665                    SFC Architecture                October 2015


1.4.  Definition of Terms



   Network Service:  An offering provided by an operator that is
        delivered using one or more service functions.  This may also be
        referred to as a "composite service".  The term "service" is
        used to denote a "network service" in the context of this
        document.

        Note: Beyond this document, the term "service" is overloaded
        with varying definitions.  For example, to some a service is an
        offering composed of several elements within the operator's
        network, whereas for others a service, or more specifically a
        network service, is a discrete element such as a "firewall".
        Traditionally, such services (in the latter sense) host a set of
        service functions and have a network locator where the service
        is hosted.

   Classification:  Locally instantiated matching of traffic flows
        against policy for subsequent application of the required set of
        network service functions.  The policy may be customer/network/
        service specific.

   Classifier:  An element that performs Classification.

   Service Function Chain (SFC):  A service function chain defines an
        ordered set of abstract service functions and ordering
        constraints that must be applied to packets and/or frames and/or
        flows selected as a result of classification.  An example of an
        abstract service function is "a firewall".  The implied order
        may not be a linear progression as the architecture allows for
        SFCs that copy to more than one branch, and also allows for
        cases where there is flexibility in the order in which service
        functions need to be applied.  The term "service chain" is often
        used as shorthand for service function chain.

   Service Function (SF):  A function that is responsible for specific
        treatment of received packets.  A Service Function can act at
        various layers of a protocol stack (e.g., at the network layer
        or other OSI layers).  As a logical component, a service
        function can be realized as a virtual element or be embedded in
        a physical network element.  One or more Service Functions can
        be embedded in the same network element.  Multiple occurrences
        of the service function can exist in the same administrative
        domain.

        One or more service functions can be involved in the delivery of
        added-value services.  A non-exhaustive list of abstract service
        functions includes: firewalls, WAN and application acceleration,



Halpern & Pignataro           Informational                     [Page 6]

RFC 7665                    SFC Architecture                October 2015


        Deep Packet Inspection (DPI), Lawful Intercept (LI), server load
        balancing, NAT44 [RFC3022], NAT64 [RFC6146], NPTv6 [RFC6296],
        HOST_ID injection, HTTP Header Enrichment functions, and TCP
        optimizer.

        An SF may be SFC encapsulation aware (that is, it receives and
        acts on information in the SFC encapsulation) or unaware (in
        which case, data forwarded to the SF does not contain the SFC
        encapsulation).  This is often referred to as "SFC aware" and
        "SFC unaware", respectively.

   Service Function Forwarder (SFF):  A service function forwarder is
        responsible for forwarding traffic to one or more connected
        service functions according to information carried in the SFC
        encapsulation, as well as handling traffic coming back from the
        SF.  Additionally, an SFF is responsible for delivering traffic
        to a classifier when needed and supported, transporting traffic
        to another SFF (in the same or different type of overlay), and
        terminating the Service Function Path (SFP).

   Metadata:  Provides the ability to exchange context information
        between classifiers and SFs, and among SFs.

   Service Function Path (SFP):  The service function path is a
        constrained specification of where packets assigned to a certain
        service function path must go.  While it may be so constrained
        as to identify the exact locations, it can also be less
        specific.  The SFP provides a level of indirection between the
        fully abstract notion of service chain as a sequence of abstract
        service functions to be delivered, and the fully specified
        notion of exactly which SFF/SFs the packet will visit when it
        actually traverses the network.  By allowing the control
        components to specify this level of indirection, the operator
        may control the degree of SFF/SF selection authority that is
        delegated to the network.

   SFC Encapsulation:  The SFC encapsulation provides, at a minimum, SFP
        identification, and is used by the SFC-aware functions, such as
        the SFF and SFC-aware SFs.  The SFC encapsulation is not used
        for network packet forwarding.  In addition to SFP
        identification, the SFC encapsulation carries metadata including
        data-plane context information.









Halpern & Pignataro           Informational                     [Page 7]

RFC 7665                    SFC Architecture                October 2015


   Rendered Service Path (RSP):  Within an SFP, packets themselves are
        of course transmitted from and to specific places in the
        network, visiting a specific sequence of SFFs and SFs.  This
        sequence of actual visits by a packet to specific SFFs and SFs
        in the network is known as the Rendered Service Path (RSP).
        This definition is included here for use by later documents,
        such as when solutions may need to discuss the actual sequence
        of locations the packets visit.

   SFC-Enabled Domain:  A network or region of a network that implements
        SFC.  An SFC-enabled domain is limited to a single network
        administrative domain.

   SFC Proxy:  Removes and inserts SFC encapsulation on behalf of an
        SFC-unaware service function.  SFC proxies are logical elements.

2.  Architectural Concepts



   The following sections describe the foundational concepts of service
   function chaining and the SFC architecture.

   Service function chaining enables the creation of composite (network)
   services that consist of an ordered set of SFs that must be applied
   to packets and/or frames and/or flows selected as a result of
   classification.  Each SF is referenced using an identifier that is
   unique within an SF-enabled domain.

   Service function chaining is a concept that provides for more than
   just the application of an ordered set of SFs to selected traffic;
   rather, it describes a method for deploying SFs in a way that enables
   dynamic ordering and topological independence of those SFs as well as
   the exchange of metadata between participating entities.

2.1.  Service Function Chains



   In most networks, services are constructed as abstract sequences of
   SFs that represent SFCs.  At a high level, an SFC is an abstracted
   view of a service that specifies the set of required SFs as well as
   the order in which they must be executed.  Graphs, as illustrated in
   Figure 1, define an SFC, where each graph node represents the
   required existence of at least one abstract SF.  Such graph nodes
   (SFs) can be part of zero, one, or many SFCs.  A given graph node
   (SF) can appear one time or multiple times in a given SFC.

   SFCs can start from the origination point of the service function
   graph (i.e., node 1 in Figure 1), or from any subsequent node in the
   graph.  As shown, SFs may therefore become branching nodes in the
   graph, with those SFs selecting edges that move traffic to one or



Halpern & Pignataro           Informational                     [Page 8]

RFC 7665                    SFC Architecture                October 2015


   more branches.  The top and middle graphs depict such a case, where a
   second classification event occurs after node 2, and a new graph is
   selected (i.e., node 3 instead of node 6).  The bottom graph
   highlights the concept of a cycle, in which a given SF (e.g., node 7
   in the depiction) can be visited more than once within a given
   service chain.  An SFC can have more than one terminus.

     ,-+-.         ,---.          ,---.          ,---.
    /     \       /     \        /     \        /     \
   (   1   )+--->(   2   )+---->(   6   )+---->(   8   )
    \     /       \     /        \     /        \     /
     `---'         `---'          `---'          `---'

     ,-+-.         ,---.          ,---.          ,---.          ,---.
    /     \       /     \        /     \        /     \        /     \
   (   1   )+--->(   2   )+---->(   3   )+---->(   7   )+---->(   9   )
    \     /       \     /        \     /        \     /        \     /
     `---'         `---'          `---'          `---'          `---'

     ,-+-.         ,---.          ,---.          ,---.          ,---.
    /     \       /     \        /     \        /     \        /     \
   (   1   )+--->(   7   )+---->(   8   )+---->(   4   )+---->(   7   )
    \     /       \     /        \     /        \     /        \     /
     `---'         `---'          `---'          `---'          `---'

                  Figure 1: Service Function Chain Graphs

   The concepts of classification, reclassification, and branching are
   covered in subsequent sections of this architecture (see Sections 4.7
   and 4.8).

2.2.  Service Function Chain Symmetry



   SFCs may be unidirectional or bidirectional.  A unidirectional SFC
   requires that traffic be forwarded through the ordered SFs in one
   direction (sf1 -> sf2 -> sf3), whereas a bidirectional SFC requires a
   symmetric path (sf1 -> sf2 -> sf3 and sf3 -> sf2 -> sf1), and in
   which the SF instances are the same in opposite directions.  A hybrid
   SFC has attributes of both unidirectional and bidirectional SFCs;
   that is to say some SFs require symmetric traffic, whereas other SFs
   do not process reverse traffic or are independent of the
   corresponding forward traffic.

   SFCs may contain cycles; that is traffic may need to traverse one or
   more SFs within an SFC more than once.  Solutions will need to ensure
   suitable disambiguation for such situations.





Halpern & Pignataro           Informational                     [Page 9]

RFC 7665                    SFC Architecture                October 2015


   The architectural allowance that is made for SFPs that delegate
   choice to the network for which SFs and/or SFFs a packet will visit
   creates potential issues here.  A solution that allows such
   delegation needs to also describe how the solution ensures that those
   service chains requiring service function chain symmetry can achieve
   that.

   Further, there are state trade-offs in symmetry.  Symmetry may be
   realized in several ways depending on the SFF and classifier
   functionality.  In some cases, "mirrored" classification (i.e., from
   Source to Destination and from Destination to Source) policy may be
   deployed, whereas in others shared state between classifiers may be
   used to ensure that symmetric flows are correctly identified, then
   steered along the required SFP.  At a high level, there are various
   common cases.  In a non-exhaustive way, there can be for example:

   o  A single classifier (or a small number of classifiers), in which
      case both incoming and outgoing flows could be recognized at the
      same classifier, so the synchronization would be feasible by
      internal mechanisms internal to the classifier.

   o  Stateful classifiers where several classifiers may be clustered
      and share state.

   o  Fully distributed classifiers, where synchronization needs to be
      provided through unspecified means.

   o  A classifier that learns state from the egress packets/flows that
      is then used to provide state for the return packets/flow.

   o  Symmetry may also be provided by stateful forwarding logic in the
      SFF in some implementations.

   This is a non-comprehensive list of common cases.

2.3.  Service Function Paths



   A Service Function Path (SFP) is a mechanism used by service chaining
   to express the result of applying more granular policy and
   operational constraints to the abstract requirements of a service
   chain (SFC).  This architecture does not mandate the degree of
   specificity of the SFP.  Architecturally, within the same SFC-enabled
   domain, some SFPs may be fully specified, selecting exactly which SFF
   and which SF are to be visited by packets using that SFP, while other
   SFPs may be quite vague, deferring to the SFF the decisions about the
   exact sequence of steps to be used to realize the SFC.  The
   specificity may be anywhere in between these extremes.




Halpern & Pignataro           Informational                    [Page 10]

RFC 7665                    SFC Architecture                October 2015


   As an example of such an intermediate specificity, there may be two
   SFPs associated with a given SFC, where one SFP specifies that any
   order of SFF and SF may be used as long as it is within Data Center
   1, and where the second SFP allows the same latitude, but only within
   Data Center 2.

   Thus, the policies and logic of SFP selection or creation (depending
   upon the solution) produce what may be thought of as a constrained
   version of the original SFC.  Since multiple policies may apply to
   different traffic that uses the same SFC, it also follows that there
   may be multiple SFPs associated with a single SFC.

   The architecture allows for the same SF to be reachable through
   multiple SFFs.  In these cases, some SFPs may constrain which SFF is
   used to reach which SF, while some SFPs may leave that decision to
   the SFF itself.

   Further, the architecture allows for two or more SFs to be attached
   to the same SFF, and possibly connected via internal means allowing
   more effective communication.  In these cases, some solutions or
   deployments may choose to use some form of internal inter-process or
   inter-VM messaging (communication behind the virtual switching
   element) that is optimized for such an environment.  This must be
   coordinated with the SFF so that it can properly perform its job.
   Implementation details of such mechanisms are considered out of scope
   for this document, and can include a spectrum of methods: for
   example, situations including all next-hops explicitly, others where
   a list of possible next-hops is provided and the selection is local,
   or cases with just an identifier, where all resolution is local.

   This architecture also allows the same SF to be part of multiple
   SFPs.

2.3.1.  Service Function Chains, Service Function Paths, and Rendered
        Service Path



   As an example of this progressive refinement, consider a Service
   Function Chain (SFC) that states that packets using this chain should
   be delivered to a firewall and a caching engine.

   A Service Function Path (SFP) could refine this, considering that
   this architecture does not mandate the degree of specificity an SFP
   has to have.  It might specify that the firewall and caching engine
   are both to be in a specific data center (e.g., in DC1), or it might
   specify exactly which instance of each firewall and caching engine is
   to be used.





Halpern & Pignataro           Informational                    [Page 11]

RFC 7665                    SFC Architecture                October 2015


   The Rendered Service Path (RSP) is the actual sequence of SFFs and
   SFs that the packets will actually visit.  So if the SFP picked the
   DC, the RSP would be more specific.

3.  Architecture Principles



   Service function chaining is predicated on several key architectural
   principles:

   1.  Topological independence: No changes to the underlay network
       forwarding topology -- implicit, or explicit -- are needed to
       deploy and invoke SFs or SFCs.

   2.  Plane separation: Dynamic realization of SFPs is separated from
       packet handling operations (e.g., packet forwarding).

   3.  Classification: Traffic that satisfies classification rules is
       forwarded according to a specific SFP.  For example,
       classification can be as simple as an explicit forwarding entry
       that forwards all traffic from one address into the SFP.
       Multiple classification points are possible within an SFC (i.e.,
       forming a service graph), thus enabling changes/updates to the
       SFC by SFs.

       Classification can occur at varying degrees of granularity; for
       example, classification can use a 5-tuple, a transport port or
       set of ports, part of the packet payload, it can be the result of
       high-level inspections, or it can come from external systems.

   4.  Shared Metadata: Metadata/context data can be shared amongst SFs
       and classifiers, between SFs, and between external systems and
       SFs (e.g., orchestration).

       One use of metadata is to provide and share the result of
       classification (that occurs within the SFC-enabled domain, or
       external to it) along an SFP.  For example, an external
       repository might provide user/subscriber information to a service
       chain classifier.  This classifier could in turn impose that
       information in the SFC encapsulation for delivery to the
       requisite SFs.  The SFs could in turn utilize the user/subscriber
       information for local policy decisions.  Metadata can also share
       SF output along the SFP.

   5.  Service definition independence: The SFC architecture does not
       depend on the details of SFs themselves.








Halpern & Pignataro           Informational                    [Page 12]

RFC 7665                    SFC Architecture                October 2015


   6.  Service function chain independence: The creation, modification,
       or deletion of an SFC has no impact on other SFCs.  The same is
       true for SFPs.



   7.  Heterogeneous control/policy points: The architecture allows SFs
       to use independent mechanisms (out of scope for this document) to
       populate and resolve local policy and (if needed) local
       classification criteria.

4.  Core SFC Architecture Components



   The SFC Architecture is built out of architectural building blocks
   that are logical components; these logical components are
   classifiers, Service Function Forwarders (SFFs), the Service
   Functions (SFs) themselves, and SFC proxies.  While this architecture
   describes functionally distinct logical components and promotes
   transport independence, they could be realized and combined in
   various ways in deployed products, and could be combined with an
   overlay.

   They are interconnected using the SFC encapsulation.  This results in
   a high-level logical architecture of an SFC-enabled domain that
   comprises:

      o . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
      .  +--------------+                  +------------------~~~
      .  |   Service    |       SFC        |  Service  +---+   +---+
      .  |Classification|  Encapsulation   | Function  |sf1|...|sfn|
   +---->|   Function   |+---------------->|   Path    +---+   +---+
      .  +--------------+                  +------------------~~~
      . SFC-enabled Domain
      o . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

               Figure 2: Service Function Chain Architecture

   The following subsections provide details on each logical component
   that form the basis of the SFC architecture.  A detailed overview of
   how some of these architectural components interact is provided in
   Figure 3:












Halpern & Pignataro           Informational                    [Page 13]

RFC 7665                    SFC Architecture                October 2015


          +----------------+                        +----------------+
          |   SFC-aware    |                        |  SFC-unaware   |
          |Service Function|                        |Service Function|
          +-------+--------+                        +-------+--------+
                  |                                         |
            SFC Encapsulation                       No SFC Encapsulation
                  |                  SFC                    |
     +---------+  +----------------+ Encapsulation     +---------+
     |SFC-Aware|-----------------+  \     +------------|SFC Proxy|
     |    SF   | ... ----------+  \  \   /             +---------+
     +---------+                \  \  \ /
                               +-------+--------+
                               |   SF Forwarder |
                               |      (SFF)     |
                               +-------+--------+
                                       |
                               SFC Encapsulation
                                       |
                           ... SFC-enabled Domain ...
                                       |
                           Network Overlay Transport
                                       |
                                   _,....._
                                ,-'        `-.
                               /              `.
                              |     Network    |
                              `.              /
                                `.__     __,-'
                                    `''''

    Figure 3: SFC Architecture Components After Initial Classification

   Please note that the depiction in Figure 3 shows packets after
   initial classification, and therefore includes the SFC encapsulation.
   Although not included in Figure 3, the classifier is an SFC
   architectural component.

4.1.  SFC Encapsulation



   The SFC encapsulation enables service function path selection.  It
   also enables the sharing of metadata/context information when such
   metadata exchange is required.

   The SFC encapsulation carries explicit information used to identify
   the SFP.  However, the SFC encapsulation is not a transport
   encapsulation itself: it is not used to forward packets within the
   network fabric.  If packets need to flow between separate physical
   platforms, the SFC encapsulation relies on an outer network



Halpern & Pignataro           Informational                    [Page 14]

RFC 7665                    SFC Architecture                October 2015


   transport.  Transit forwarders -- such as router and switches --
   forward SFC encapsulated packets based on the outer (non-SFC)
   encapsulation.

   One of the key architecture principles of SFC is that the SFC
   encapsulation remain transport independent.  As such, any network
   transport protocol may be used to carry the SFC encapsulated traffic.

4.2.  Service Function (SF)



   The concept of an SF evolves; rather than being viewed as a bump in
   the wire, an SF becomes a resource within a specified administrative
   domain that is available for consumption as part of a composite
   service.  SFs send/receive data to/from one or more SFFs.  SFC-aware
   SFs receive this traffic with the SFC encapsulation.

   While the SFC architecture defines the concept and specifies some
   characteristics of a new encapsulation -- the SFC encapsulation --
   and several logical components for the construction of SFCs, existing
   SF implementations may not have the capabilities to act upon or fully
   integrate with the new SFC encapsulation.  In order to provide a
   mechanism for such SFs to participate in the architecture, an SFC
   proxy function is defined (see Section 4.6).  The SFC proxy acts as a
   gateway between the SFC encapsulation and SFC-unaware SFs.  The
   integration of SFC-unaware service functions is discussed in more
   detail in the SFC proxy section.

   This architecture allows an SF to be part of multiple SFPs and SFCs.

4.3.  Service Function Forwarder (SFF)



   The SFF is responsible for forwarding packets and/or frames received
   from the network to one or more SFs associated with a given SFF using
   information conveyed in the SFC encapsulation.  Traffic from SFs
   eventually returns to the same SFF, which is responsible for
   injecting traffic back onto the network.  Some SFs, such as
   firewalls, could also consume a packet.

   The collection of SFFs and associated SFs creates a service-plane
   overlay in which SFC-aware SFs, as well as SFC-unaware SFs reside.
   Within this service plane, the SFF component connects different SFs
   that form a service function path.

   SFFs maintain the requisite SFP forwarding information.  SFP
   forwarding information is associated with a service path identifier
   that is used to uniquely identify an SFP.  The service forwarding
   state enables an SFF to identify which SFs of a given SFP should be
   applied, and in what order, as traffic flows through the associated



Halpern & Pignataro           Informational                    [Page 15]

RFC 7665                    SFC Architecture                October 2015


   SFP.  While there may appear to the SFF to be only one available way
   to deliver the given SF, there may also be multiple choices allowed
   by the constraints of the SFP.

   If there are multiple choices, the SFF needs to preserve the property
   that all packets of a given flow are handled the same way, since the
   SF may well be stateful.  Additionally, the SFF may preserve the
   handling of packets based on other properties on top of a flow, such
   as a subscriber, session, or application instance identification.

   The SFF also has the information that allows it to forward packets to
   the next SFF after applying local service functions.  Again, while
   there may be only a single choice available, the architecture allows
   for multiple choices for the next SFF.  As with SFs, the solution
   needs to operate such that the behavior with regard to specific flows
   (see the Rendered Service Path) is stable.  The selection of
   available SFs and next SFFs may be interwoven when an SFF supports
   multiple distinct service functions and the same service function is
   available at multiple SFFs.  Solutions need to be clear about what is
   allowed in these cases.

   Even when the SFF supports and utilizes multiple choices, the
   decision as to whether to use flow-specific mechanisms or coarser-
   grained means to ensure that the behavior of specific flows is stable
   is a matter for specific solutions and specific implementations.

   The SFF component has the following primary responsibilities:

   1.  SFP forwarding: Traffic arrives at an SFF from the network.  The
       SFF determines the appropriate SF the traffic should be forwarded
       to via information contained in the SFC encapsulation.  After SF
       processing, the traffic is returned to the SFF, and, if needed,
       is forwarded to another SF associated with that SFF.  If there is
       another non-local (i.e., different SFF) hop in the SFP, the SFF
       further encapsulates the traffic in the appropriate network
       transport protocol and delivers it to the network for delivery to
       the next SFF along the path.  Related to this forwarding
       responsibility, an SFF should be able to interact with metadata.

   2.  Terminating SFPs: An SFC is completely executed when traffic has
       traversed all required SFs in a chain.  When traffic arrives at
       the SFF after the last SF has finished processing it, the final
       SFF knows from the service forwarding state that the SFC is
       complete.  The SFF removes the SFC encapsulation and delivers the
       packet back to the network for forwarding.






Halpern & Pignataro           Informational                    [Page 16]

RFC 7665                    SFC Architecture                October 2015


   3.  Maintaining flow state: In some cases, the SFF may be stateful.
       It creates flows and stores flow-centric information.  This state
       information may be used for a range of SFP-related tasks such as
       ensuring consistent treatment of all packets in a given flow,
       ensuring symmetry, or for state-aware SFC Proxy functionality
       (see Section 4.8).

4.3.1.  Transport-Derived SFF



   SFP forwarding, as described above, directly depends upon the use of
   the service path information contained in the SFC encapsulation.
   However, existing implementations may not be able to act on the SFC
   encapsulation.  These platforms may opt to use existing transport
   information, if it can be arranged, to provide explicit service path
   information.

   This results in the same architectural behavior and meaning for SFP
   forwarding and service function paths.  It is the responsibility of
   the control components to ensure that the transport path executed in
   such a case is fully aligned with the path identified by the
   information in the service chaining encapsulation.

4.4.  SFC-Enabled Domain



   Specific features may need to be enforced at the boundaries of an
   SFC-enabled domain, for example to avoid leaking SFC information.
   Using the term "node" to refer generically to an entity that is
   performing a set of functions, in this context, an SFC boundary node
   denotes a node that connects one SFC-enabled domain to a node either
   located in another SFC-enabled domain or in a domain that is SFC-
   unaware.

   An SFC boundary node can act as egress or ingress.  An SFC Egress
   Node denotes an SFC boundary node that handles traffic leaving the
   SFC-enabled domain the Egress Node belongs to.  Such a node is
   required to remove any information specific to the SFC Domain,
   typically the SFC encapsulation.  Further, from a privacy
   perspective, an SFC Egress Node is required to ensure that any
   sensitive information added as part of SFC gets removed.  In this
   context, information may be sensitive due to network concerns or end-
   customer concerns.  An SFC Ingress Node denotes an SFC boundary node
   that handles traffic entering the SFC-enabled domain.  In most
   solutions and deployments this will need to include a classifier, and
   will be responsible for adding the SFC encapsulation to the packet.

   An SFC Proxy and corresponding SFC-unaware service function (see
   Figure 3) are inside the SFC-enabled domain.




Halpern & Pignataro           Informational                    [Page 17]

RFC 7665                    SFC Architecture                October 2015


4.5.  Network Overlay and Network Components



   Underneath the SFF there are components responsible for performing
   the transport (overlay) forwarding.  They do not consult the SFC
   encapsulation or inner payload for performing this forwarding.  They
   only consult the outer-transport encapsulation for the transport
   (overlay) forwarding.

4.6.  SFC Proxy



   In order for the SFC architecture to support SFC-unaware SFs (e.g.,
   legacy service functions) a logical SFC proxy function may be used.
   This function sits between an SFF and one or more SFs to which the
   SFF is directing traffic (see Figure 3).

   The proxy accepts packets from the SFF on behalf of the SF.  It
   removes the SFC encapsulation, and then uses a local attachment
   circuit to deliver packets to SFC-unaware SFs.  It also receives
   packets back from the SF, reapplies the SFC encapsulation, and
   returns them to the SFF for processing along the service function
   path.

   Thus, from the point of view of the SFF, the SFC proxy appears to be
   part of an SFC-aware SF.

   Communication details between the SFF and the SFC Proxy are the same
   as those between the SFF and an SFC-aware SF.  The details of that
   are not part of this architecture.  The details of the communication
   methods over the local attachment circuit between the SFC proxy and
   the SFC-unaware SF are dependent upon the specific behaviors and
   capabilities of that SFC-unaware SF, and thus are also out of scope
   for this architecture.

   Specifically, for traffic received from the SFF intended for the SF
   the proxy is representing, the SFC proxy:

   o  Removes the SFC encapsulation from SFC encapsulated packets.

   o  Identifies the required SF to be applied based on available
      information including that carried in the SFC encapsulation.

   o  Selects the appropriate outbound local attachment circuit through
      which the next SF for this SFP is reachable.  This is derived from
      the identification of the SF carried in the SFC encapsulation, and
      may include local techniques.  Examples of a local attachment
      circuit include, but are not limited to, VLAN, IP-in-IP, Layer 2





Halpern & Pignataro           Informational                    [Page 18]

RFC 7665                    SFC Architecture                October 2015


      Tunneling Protocol version 3 (L2TPv3), Generic Routing
      Encapsulation (GRE), and Virtual eXtensible Local Area Network
      (VXLAN).

   o  Forwards the original payload via the selected local attachment
      circuit to the appropriate SF.

   When traffic is returned from the SF:

   o  Applies the required SFC encapsulation.  The determination of the
      encapsulation details may be inferred by the local attachment
      circuit through which the packet and/or frame was received, or via
      packet classification, or other local policy.  In some cases,
      packet ordering or modification by the SF may necessitate
      additional classification in order to reapply the correct SFC
      encapsulation.

   o  Delivers the packet with the SFC encapsulation to the SFF, as
      would happen with packets returned from an SFC-aware SF.

4.7.  Classification



   Traffic from the network that satisfies classification criteria is
   directed into an SFP and forwarded to the requisite service
   function(s).  Classification is handled by a service classification
   function; initial classification occurs at the ingress to the SFC
   domain.  The granularity of the initial classification is determined
   by the capabilities of the classifier and the requirements of the SFC
   policy.  For instance, classification might be relatively coarse: all
   packets from this port are subject to SFC policy X and directed into
   SFP A, or quite granular: all packets matching this 5-tuple are
   subject to SFC policy Y and directed into SFP B.

   As a consequence of the classification decision, the appropriate SFC
   encapsulation is imposed on the data, and a suitable SFP is selected
   or created.  Classification results in attaching the traffic to a
   specific SFP.

4.8.  Reclassification and Branching



   The SFC architecture supports reclassification (or non-initial
   classification) as well.  As packets traverse an SFP,
   reclassification may occur -- typically performed by a classification
   function co-resident with a service function.  Reclassification may
   result in the selection of a new SFP, an update of the associated
   metadata, or both.  This is referred to as "branching".





Halpern & Pignataro           Informational                    [Page 19]

RFC 7665                    SFC Architecture                October 2015


   For example, an initial classification results in the selection of
   SFP A: DPI_1 --> SLB_8.  However, when the DPI service function is
   executed, attack traffic is detected at the application layer.  DPI_1
   reclassifies the traffic as attack and alters the service path to SFP
   B, to include a firewall for policy enforcement: dropping the
   traffic: DPI_1 --> FW_4.  Subsequent to FW_4, surviving traffic would
   be returned to the original SFF.  In this simple example, the DPI
   service function reclassifies the traffic based on local application
   layer classification capabilities (that were not available during the
   initial classification step).

   When traffic arrives after being steered through an SFC-unaware SF,
   the SFC Proxy must perform reclassification of traffic to determine
   the SFP.  The SFC Proxy is concerned with re-attaching information
   for SFC-unaware SFs, and a stateful SFC Proxy simplifies such
   classification to a flow lookup.

4.9.  Shared Metadata



   Sharing metadata allows the network to provide network-derived
   information to the SFs, SF-to-SF information exchange, and the
   sharing of service-derived information to the network.  Some SFCs may
   not require metadata exchange.  SFC infrastructure enables the
   exchange of this shared data along the SFP.  The shared metadata
   serves several possible roles within the SFC architecture:

   o  Allows elements that typically operate independently (e.g., as
      "ships in the night") to exchange information.

   o  Encodes information about the network and/or data for subsequent
      use within the SFP.

   o  Creates an identifier used for policy binding by SFs.

   Context information can be derived in several ways:

   o  External sources

   o  Network node classification

   o  Service function classification










Halpern & Pignataro           Informational                    [Page 20]

RFC 7665                    SFC Architecture                October 2015


5.  Additional Architectural Concepts

   There are a number of issues that solutions need to address, and that
   the architecture informs but does not determine.  This section lays
   out some of those concepts.

5.1.  The Role of Policy



   Much of the behavior of service chains is driven by operator and per-
   customer policy.  This architecture is structured to isolate the
   policy interactions from the data plane and control logic.

   Specifically, it is assumed that the service chaining control plane
   creates the service paths.  The service chaining data plane is used
   to deliver the classified packets along the service chains to the
   intended service functions.

   Policy, in contrast, interacts with the system in other places.
   Policies and policy engines may monitor service functions to decide
   if additional (or fewer) instances of services are needed.  When
   applicable, those decisions may in turn result in interactions that
   direct the control logic to change the SFP placement or packet
   classification rules.

   Similarly, operator service policy, often managed by Operations or
   Business Support Systems (OSS or BSS), will frequently determine what
   service functions are available.  Operator service policies also
   determine which sequences of functions are valid and are to be used
   or made available.

   The offering of service chains to customers, and the selection of
   which service chain a customer wishes to use, are driven by a
   combination of operator and customer policies using appropriate
   portals in conjunction with the OSS and BSS tools.  These selections
   then drive the service chaining control logic, which in turn
   establishes the appropriate packet classification rules.

5.2.  SFC Control Plane



   The SFC control plane is part of the overall SFC architecture, and
   this section describes its high-level functions.  However, the
   detailed definition of the SFC control plane is outside the scope of
   this document.

   The SFC control plane is responsible for constructing SFPs,
   translating SFCs to forwarding paths, and propagating path
   information to participating nodes to achieve requisite forwarding
   behavior to construct the service overlay.  For instance, an SFC



Halpern & Pignataro           Informational                    [Page 21]

RFC 7665                    SFC Architecture                October 2015


   construction may be static; selecting exactly which SFFs and which
   SFs from those SFFs are to be used, or it may be dynamic, allowing
   the network to perform some or all of the choices of SFF or SF to use
   to deliver the selected service chain within the constraints
   represented by the service path.

   In the SFC architecture, SFs are resources; the control plane manages
   and communicates their capabilities, availability, and location in
   fashions suitable for the transport and SFC operations in use.  The
   control plane is also responsible for the creation of the context
   (see below).  The control plane may be distributed (using new or
   existing control-plane protocols), or be centralized, or a
   combination of the two.

   The SFC control plane provides the following functionality:

   1.  An SFC-enabled domain wide view of all available service function
       resources as well as the network locators through which they are
       reachable.

   2.  Uses SFC policy to construct service function chains, and
       associated SFPs.

   3.  Selection of specific SFs for a requested SFC, either statically
       (using specific SFs) or dynamically (using service explicit SFs
       at the time of delivering traffic to them).

   4.  Provides requisite SFC data-plane information to the SFC
       architecture components, most notably the SFF.

   5.  Provides the metadata and usage information classifiers need so
       that they in turn can provide this metadata for appropriate
       packets in the data plane.

   6.  When needed, provide information including policy information to
       other SFC elements to be able to properly interpret metadata.

5.3.  Resource Control



   The SFC system may be responsible for managing all resources
   necessary for the SFC components to function.  This includes network
   constraints used to plan and choose network path(s) between service
   function forwarders, network communication paths between service
   function forwarders and their attached service functions,
   characteristics of the nodes themselves such as memory, number of
   virtual interfaces, routes, and instantiation, configuration, and
   deletion of SFs.




Halpern & Pignataro           Informational                    [Page 22]

RFC 7665                    SFC Architecture                October 2015


   The SFC system will also be required to reflect policy decisions
   about resource control, as expressed by other components in the
   system.

   While all of these aspects are part of the overall system, they are
   beyond the scope of this architecture.

5.4.  Infinite Loop Detection and Avoidance



   This SFC architecture is predicated on topological independence from
   the underlying forwarding topology.  Consequently, a service topology
   is created by service function paths or by the local decisions of the
   service function forwarders based on the constraints expressed in the
   SFP.  Due to the overlay constraints, the packet-forwarding path may
   need to visit the same SFF multiple times, and in some less common
   cases may even need to visit the same SF more than once.  The Service
   Chaining solution needs to permit these limited and policy-compliant
   loops.  At the same time, the solutions must ensure that indefinite
   and unbounded loops cannot be formed, as such would consume unbounded
   resources without delivering any value.

   In other words, this architecture requires the solution to prevent
   infinite service function loops, even when service functions may be
   invoked multiple times in the same SFP.

5.5.  Load-Balancing Considerations



   Supporting function elasticity and high-availability should not
   overly complicate SFC or lead to unnecessary scalability problems.

   In the simplest case, where there is only a single function in the
   SFP (the next hop is either the destination address of the flow or
   the appropriate next hop to that destination), one could argue that
   there may be no need for SFC.

   In the cases where the classifier is separate from the single
   function or a function at the terminal address may need a sub-prefix
   (e.g., finer-grained address information) or per-subscriber metadata,
   a single SFP exists (i.e., the metadata changes but the SFP does
   not), regardless of the number of potential terminal addresses for
   the flow.  This is the case of the simple load balancer.  See
   Figure 4.

                            +---+    +---++--->web server
                  source+-->|sff|+-->|sf1|+--->web server
                            +---+    +---++--->web server

                      Figure 4: Simple Load Balancing



Halpern & Pignataro           Informational                    [Page 23]

RFC 7665                    SFC Architecture                October 2015


   By extrapolation, in the case where intermediary functions within a
   chain had similar "elastic" behaviors, we do not need separate chains
   to account for this behavior -- as long as the traffic coalesces to a
   common next-hop after the point of elasticity.

   In Figure 5, we have a chain of five service functions between the
   traffic source and its destination.

                +---+ +---+ +---+   +---+ +---+ +---+
                |sf2| |sf2| |sf3|   |sf3| |sf4| |sf4|
                +---+ +---+ +---+   +---+ +---+ +---+
                  |     |     |       |     |     |
                  +-----+-----+       +-----+-----+
                        |                   |
                        +                   +
             +---+    +---+     +---+     +---+    +---+
   source+-->|sff|+-->|sff|+--->|sff|+--->|sff|+-->|sff|+-->destination
             +---+    +---+     +---+     +---+    +---+
               +                  +                  +
               |                  |                  |
             +---+              +---+              +---+
             |sf1|              |sf3|              |sf5|
             +---+              +---+              +---+

                         Figure 5: Load Balancing

   This would be represented as one service function path:
   sf1 -> sf2 -> sf3 -> sf4 -> sf5.  The SFF is a logical element, which
   may be made up of one or multiple components.  In this architecture,
   the SFF may handle load distribution based on policy.

   It can also be seen in the above that the same service function may
   be reachable through multiple SFFs, as discussed earlier.  The
   selection of which SFF to use to reach sf3 may be made by the control
   logic in defining the SFP, or may be left to the SFFs themselves,
   depending upon policy, solution, and deployment constraints.  In the
   latter case, it needs to be assured that exactly one SFF takes
   responsibility to steer traffic through sf3.

5.6.  MTU and Fragmentation Considerations



   This architecture prescribes that additional information be added to
   packets to identify service function paths and often to represent
   metadata.  It also envisions adding transport information to carry
   packets along service function paths, at least between service
   function forwarders.  This added information increases the size of
   the packet to be carried by service chaining.  Such additions could




Halpern & Pignataro           Informational                    [Page 24]

RFC 7665                    SFC Architecture                October 2015


   potentially increase the packet size beyond the MTU supported on some
   or all of the media used in the service chaining domain.

   Such packet size increases can thus cause operational MTU problems.
   Requiring fragmentation and reassembly in an SFF would be a major
   processing increase and might be impossible with some transports.
   Expecting service functions to deal with packets fragmented by the
   SFC function might be onerous even when such fragmentation was
   possible.  Thus, at the very least, solutions need to pay attention
   to the size cost of their approach.  There may be alternative or
   additional means available, although any solution needs to consider
   the trade-offs.

   These considerations apply to any generic architecture that increases
   the header size.  There are also more specific MTU considerations:
   Effects on Path MTU Discovery (PMTUD) as well as deployment
   considerations.  Deployments within a single administrative control
   or even a single data center complex can afford more flexibility in
   dealing with larger packets, and deploying existing mitigations that
   decrease the likelihood of fragmentation or discard.

5.7.  SFC OAM



   Operations, Administration, and Maintenance (OAM) tools are an
   integral part of the architecture.  These serve various purposes,
   including fault detection and isolation, and performance management.
   For example, there are many advantages of SFP liveness detection,
   including status reporting, support for resiliency operations and
   policies, and an enhanced ability to balance load.

   Service function paths create a services topology, and OAM performs
   various functions within this service layer.  Furthermore, SFC OAM
   follows the same architectural principles of SFC in general.  For
   example, topological independence (including the ability to run OAM
   over various overlay technologies) and classification-based policy.

   We can subdivide the SFC OAM architecture in two parts:

   o  In-band: OAM packets follow the same path and share fate with user
      packets, within the service topology.  For this, they also follow
      the architectural principle of consistent policy identifiers, and
      use the same path IDs as the service chain data packets.  Load
      balancing and SFC encapsulation with packet forwarding are
      particularly important here.

   o  Out-of-band: Reporting beyond the actual data plane.  An
      additional layer beyond the data-plane OAM allows for additional
      alerting and measurements.



Halpern & Pignataro           Informational                    [Page 25]

RFC 7665                    SFC Architecture                October 2015


   This architecture prescribes end-to-end SFP OAM functions, which
   implies SFF understanding of whether an in-band packet is an OAM or
   user packet.  However, service function validation is outside of the
   scope of this architecture, and application-level OAM is not what
   this architecture prescribes.

   Some of the detailed functions performed by SFC OAM include fault
   detection and isolation in a service function path or a service
   function, verification that connectivity using SFPs is both effective
   and directing packets to the intended service functions, service path
   tracing, diagnostic and fault isolation, alarm reporting, performance
   measurement, locking and testing of service functions, validation
   with the control plane (see Section 5.2), and also allow for vendor-
   specific as well as experimental functions.  SFC should leverage and,
   if needed, extend relevant existing OAM mechanisms.

5.8.  Resilience and Redundancy



   As a practical operational requirement, any service chaining solution
   needs to be able to respond effectively, and usually very quickly, to
   failure conditions.  These may be failures of connectivity in the
   network between SFFs, failures of SFFs, or failures of SFs.  Per-SF
   state (as, for example, stateful-firewall state) is the
   responsibility of the SF, and not addressed by this architecture.

   Multiple techniques are available to address this issue.  Solutions
   can describe both what they require and what they allow to address
   failure.  Solutions can make use of flexible specificity of service
   function paths, if the SFF can be given enough information in a
   timely fashion to do this.  Solutions can also make use of MAC- or
   IP-level redundancy mechanisms such as Virtual Router Redundancy
   Protocol (VRRP).  Also, particularly for SF failures, load balancers
   co-located with the SFF or as part of the service function delivery
   mechanism can provide such robustness.

   Similarly, operational requirements imply resilience in the face of
   load changes.  While mechanisms for managing (e.g., monitoring,
   instantiating, loading images, providing configuration to SFC
   control, deleting, etc.) virtual machines are out of scope for this
   architecture, solutions can and are aided by describing how they can
   make use of scaling mechanisms.










Halpern & Pignataro           Informational                    [Page 26]

RFC 7665                    SFC Architecture                October 2015


6.  Security Considerations

   The architecture described here is different from the current model,
   and moving to the new model could lead to different security
   arrangements and modeling.  In the SFC architecture, a relatively
   static topologically-dependent deployment model is replaced with the
   chaining of sets of service functions.  This can change the flow of
   data through the network, and the security and privacy considerations
   of the protocol and deployment will need to be reevaluated in light
   of the new model.

   Security considerations apply to the realization of this
   architecture, in particular to the documents that will define
   protocols.  Such realization ought to provide means to protect
   against security and privacy attacks in the areas hereby described.

   Building from the categorization of [RFC7498], we can largely divide
   the security considerations into four areas:

   Service Overlay:  Underneath the service function forwarders, the
        components that are responsible for performing the transport
        forwarding consult the outer-transport encapsulation for
        underlay forwarding.  Used transport mechanisms should satisfy
        the security requirements of the specific SFC deployment.  These
        requirements typically include varying degrees of traffic
        separation, protection against different attacks (e.g.,
        spoofing, man-in-the-middle, brute-force, or insertion attacks),
        and can also include authenticity and integrity checking, and/or
        confidentiality provisions, for both the network overlay
        transport and traffic it encapsulates.

   Boundaries:  Specific requirements may need to be enforced at the
        boundaries of an SFC-enabled domain.  These include, for
        example, to avoid leaking SFC information, and to protect its
        borders against various forms of attacks.  If untrusted parties
        can inject packets that will be treated as being properly
        classified for service chaining, there are a large range of
        attacks that can be mounted against the resulting system.
        Depending upon deployment details, these likely include spoofing
        packets from users and creating DDoS and reflection attacks of
        various kinds.  Thus, when transport mechanisms are selected for
        use with SFC, they MUST ensure that outside parties cannot
        inject SFC packets that will be accepted for processing into the
        domain.  This border security MUST include any tunnels to other
        domains.  If those tunnels are to be used for SFC without
        reclassification, then the tunnel MUST include additional
        techniques to ensure the integrity and validity of such packets.




Halpern & Pignataro           Informational                    [Page 27]

RFC 7665                    SFC Architecture                October 2015


   Classification:  Classification is used at the ingress edge of an
        SFC-enabled domain.  Policy for this classification is done
        using a plurality of methods.  Whatever method is used needs to
        consider a range of security issues.  These include appropriate
        authentication and authorization of classification policy,
        potential confidentiality issues of that policy, protection
        against corruption, and proper application of policy with needed
        segregation of application.  This includes proper controls on
        the policies that drive the application of the SFC encapsulation
        and associated metadata to packets.  Similar issues need to be
        addressed if classification is performed within a service
        chaining domain, i.e., reclassification.

   SFC Encapsulation:  The SFC encapsulation provides at a minimum SFP
        identification, and carries metadata.  An operator may consider
        the SFC Metadata as sensitive.  From a privacy perspective, a
        user may be concerned about the operator revealing data about
        (and not belonging to) the customer.  Therefore, solutions
        should consider whether there is a risk of sensitive information
        slipping out of the operator's control.  Issues of information
        exposure should also consider flow analysis.  Further, when a
        specific metadata element is defined, it should be carefully
        considered whether origin authentication is needed for it.

        A classifier may have privileged access to information about a
        packet or inside a packet (see Section 3, item 4, and
        Section 4.9) that is then communicated in the metadata.  The
        threat of leaking this private data needs to be mitigated
        [RFC6973].  As one example, if private data is represented by an
        identifier, then a new identifier can be allocated, such that
        the mapping from the private data to the new identifier is not
        broadly shared.

        Some metadata added to and carried in SFC packets is sensitive
        for various reasons, including potentially revealing personally
        identifying information.  Realizations of the architecture MUST
        protect such information to ensure that it is handled with
        suitable care and precautions against inappropriate
        dissemination.  This can have implications to the data plane,
        the control plane, or both.  Data-plane protocol definitions for
        SFC can include suitable provisions to protect such information
        for use when handling sensitive information, with packet or SFP
        granularity.  Equally, the control mechanisms used with SFC can
        have provisions to determine that such mechanisms are available,
        and to ensure that they are used when needed.  Inability to do
        so needs to result in error indications to appropriate
        management systems.  In particular, when the control systems
        know that sensitive information may potentially be added to



Halpern & Pignataro           Informational                    [Page 28]

RFC 7665                    SFC Architecture                October 2015


        packets at certain points on certain service chains, the control
        mechanism MUST verify that appropriate protective treatment of
        NSH information is available from the point where the
        information is added to the point where it will be removed.  If
        such mechanisms are unavailable, error notifications SHOULD be
        generated.

   Additionally, SFC OAM functions need to not negatively affect the
   security considerations of an SFC-enabled domain.

   Finally, all entities (software or hardware) interacting with the
   service chaining mechanisms need to provide means of security against
   malformed, poorly configured (deliberate or not) protocol constructs
   and loops.  These considerations are largely the same as those in any
   network, particularly an overlay network.

7.  References



7.1.  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>.

7.2.  Informative References



   [Boucadair2014]
              Boucadair, M., Jacquenet, C., Parker, R., Lopez, D.,
              Guichard, J., and C. Pignataro, "Service Function
              Chaining: Framework & Architecture", Work in Progress,
              draft-boucadair-sfc-framework-02, February 2014.

   [Quinn2014]
              Quinn, P. and J. Halpern, "Service Function Chaining (SFC)
              Architecture", Work in Progress, draft-quinn-sfc-arch-05,
              May 2014.

   [RFC3022]  Srisuresh, P. and K. Egevang, "Traditional IP Network
              Address Translator (Traditional NAT)", RFC 3022,
              DOI 10.17487/RFC3022, January 2001,
              <http://www.rfc-editor.org/info/rfc3022>.

   [RFC6146]  Bagnulo, M., Matthews, P., and I. van Beijnum, "Stateful
              NAT64: Network Address and Protocol Translation from IPv6
              Clients to IPv4 Servers", RFC 6146, DOI 10.17487/RFC6146,
              April 2011, <http://www.rfc-editor.org/info/rfc6146>.




Halpern & Pignataro           Informational                    [Page 29]

RFC 7665                    SFC Architecture                October 2015


   [RFC6296]  Wasserman, M. and F. Baker, "IPv6-to-IPv6 Network Prefix
              Translation", RFC 6296, DOI 10.17487/RFC6296, June 2011,
              <http://www.rfc-editor.org/info/rfc6296>.

   [RFC6973]  Cooper, A., Tschofenig, H., Aboba, B., Peterson, J.,
              Morris, J., Hansen, M., and R. Smith, "Privacy
              Considerations for Internet Protocols", RFC 6973,
              DOI 10.17487/RFC6973, July 2013,
              <http://www.rfc-editor.org/info/rfc6973>.

   [RFC7498]  Quinn, P., Ed. and T. Nadeau, Ed., "Problem Statement for
              Service Function Chaining", RFC 7498,
              DOI 10.17487/RFC7498, April 2015,
              <http://www.rfc-editor.org/info/rfc7498>.

Acknowledgments

   The editors would like to thank Sam Aldrin, Alia Atlas, Nicolas
   Bouthors, Stewart Bryant, Linda Dunbar, Alla Goldner, Ken Gray, Barry
   Greene, Anil Gunturu, David Harrington, Shunsuke Homma, Dave Hood,
   Chris Inacio, Nagendra Kumar, Hongyu Li, Andrew Malis, Guy
   Meador III, Kengo Naito, Thomas Narten, Ron Parker, Reinaldo Penno,
   Naiming Shen, Xiaohu Xu, and Lucy Yong for a thorough review and
   useful comments.

   The initial draft of this document was the result of merging two
   previous documents, and this section lists the acknowledgments from
   those documents.

   From "Service Function Chaining (SFC) Architecture" [Quinn2014]

      The authors would like to thank David Ward, Abhijit Patra, Nagaraj
      Bagepalli, Darrel Lewis, Ron Parker, Lucy Yong, and Christian
      Jacquenet for their review and comments.

   From "Service Function Chaining (SF) - Framework and Architecture"
   [Boucadair2014]:

      Many thanks to D. Abgrall, D. Minodier, Y. Le Goff, D. Cheng,
      R. White, and B. Chatras for their review and comments.











Halpern & Pignataro           Informational                    [Page 30]

RFC 7665                    SFC Architecture                October 2015


Contributors

   As noted above, this document is the result of merging two previous
   documents.  This section lists those who provided important ideas and
   text that fed into this architecture.

   The authors of "Service Function Chaining (SFC) - Framework and
   Architecture" [Boucadair2014] were:

      Mohamed Boucadair
      Christian Jacquenet
      Ron Parker
      Diego R. Lopez
      Jim Guichard
      Carlos Pignataro

   The contributors were:

      Parviz Yegani
      Paul Quinn
      Linda Dunbar

   The authors of "Service Function Chaining (SFC) Architecture"
   [Quinn2014] were:

      Paul Quinn (editor)
      Joel Halpern (editor)

   The contributors were:

      Puneet Agarwal
      Andre Beliveau
      Kevin Glavin
      Ken Gray
      Jim Guichard
      Surendra Kumar
      Darrel Lewis
      Nic Leymann
      Rajeev Manur
      Thomas Nadeau
      Carlos Pignataro
      Michael Smith
      Navindra Yadav








Halpern & Pignataro           Informational                    [Page 31]

RFC 7665                    SFC Architecture                October 2015


Authors' Addresses



   Joel Halpern (editor)
   Ericsson

   Email: jmh@joelhalpern.com


   Carlos Pignataro (editor)
   Cisco Systems, Inc.

   Email: cpignata@cisco.com







































Halpern & Pignataro           Informational                    [Page 32]