Framework for IETF Network Slices

1.1. Background

The concept of network slicing has gained traction driven largely by needs surfacing from 5G ([NGMN-NS-Concept], [TS23501], and [TS28530]). In [TS23501], a Network Slice is defined as "a logical network that provides specific network capabilities and network characteristics", and a Network Slice Instance is defined as "A set of Network Function instances and the required resources (e.g. compute, storage and networking resources) which form a deployed Network Slice." According to [TS28530], an end-to-end network slice consists of three major types of network segments: Radio Access Network (RAN), Transport Network (TN) and Core Network (CN). An IETF Network Slice provides the required connectivity between different entities in RAN and CN segments of an end-to-end network slice, with a specific performance commitment (for example, serving as a TN slice). For each end-to-end network slice, the topology and performance requirement on a customer's use of an IETF Network Slice can be very different, which requires the underlay network to have the capability of supporting multiple different IETF Network Slices.¶

While network slices are commonly discussed in the context of 5G, it is important to note that IETF Network Slices are a narrower concept with a broader usage profile, and focus primarily on particular network connectivity aspects. Other systems, including 5G deployments, may use IETF Network Slices as a component to create entire systems and concatenated constructs that match their needs, including end-to-end connectivity.¶

An IETF Network Slice could span multiple technologies and multiple administrative domains. Depending on the IETF Network Slice customer's requirements, an IETF Network Slice could be isolated from other, often concurrent IETF Network Slices in terms of data, control and management planes.¶

The customer expresses requirements for a particular IETF Network Slice service by specifying what is required rather than how the requirement is to be fulfilled. That is, the IETF Network Slice customer's view of an IETF Network Slice is an abstract one.¶

Thus, there is a need to create logical network structures with required characteristics. The customer of such a logical network can require a degree of isolation and performance that previously might not have been satisfied by overlay VPNs. Additionally, the IETF Network Slice customer might ask for some level of control of their virtual networks, e.g., to customize the service paths in a network slice.¶

This document specifies definitions and a framework for the provision of an IETF Network Slice service. Section 6 briefly indicates some candidate technologies for realizing IETF Network Slices.¶

3.1. Definition and Scope of IETF Network Slice

The term "Slice" refers to a set of characteristics and behaviors that differentiate one type of user-traffic from another within a network. An IETF Network Slice is a slice of a network that uses IETF technology. An IETF Network Slice assumes that an underlay network is capable of changing the configurations of the network devices on demand, through in-band signaling, or via controllers.¶

An IETF Network Slice enables connectivity between a set of Service Demarcation Points (SDPs) with specific Service Level Objectives (SLOs) and Service Level Expectations (SLEs) (see Section 4) over a common underlay network. Thus, an IETF Network Slice delivers a service to a customer by meeting connectivity resource requirements and associated network capabilities such as bandwidth, latency, jitter, and network functions with other resource behaviors such as compute and storage availability.¶

IETF Network Slices may be combined hierarchically, so that a network slice may itself be sliced. They may also be combined sequentially so that various different networks can each be sliced and the network slices placed into a sequence to provide an end-to-end service. This form of sequential combination is utilized in some services such as in 3GPP's 5G network [TS23501].¶

3.2. IETF Network Slice Service

A service provider delivers an IETF Network Slice service for a customer by realizing an IETF Network Slice. The IETF Network Slice service is agnostic to the technology of the underlay network, and its realization may be selected based upon multiple considerations including its service requirements and the capabilities of the underlay network. This allows an IETF Network Slice service customer to describe their network connectivity and relevant objectives in a common format, independent of the underlay technologies used.¶

The IETF Network Slice service is specified in terms of a set of SDPs, a set of one or more connectivity constructs between subsets of these SDPs, and a set of SLOs and SLEs (see Section 4) for each SDP sending to each connectivity construct. A communication type (point-to-point (P2P), point-to-multipoint (P2MP), or any-to-any (A2A)) is specified for each connectivity construct. That is, in a given IETF Network Slice service there may be one or more connectivity constructs of the same or different type, each connectivity construct may be between a different subset of SDPs, for a given connectivity construct each sending SDP has its own set of SLOs and SLEs, and the SLOs and SLEs in each set may be different. Note that a service provider may decide how many connectivity constructs per IETF Network Slice service it wishes to support such that an IETF Network Slice service may be limited to one connectivity construct or may support many.¶

This approach results in the following possible connectivity constructs:¶

• For a P2P connectivity construct, there is one sending SDP and one receiving SDP. This construct is like a private wire or a tunnel. All traffic injected at the sending SDP is intended to be received by the receiving SDP. The SLOs and SLEs apply at the sender (and implicitly at the receiver).¶
• For a P2MP connectivity construct, there is only one sending SDP and more than one receiving SDP. This is like a P2MP tunnel or multi-access VLAN segment. All traffic from the sending SDP is intended to be received by all the receiving SDPs. There is one set of SLOs and SLEs that applies at the sending SDP (and implicitly at all receiving SDPs).¶

• With an A2A connectivity construct, any sending SDP may send to any one receiving SDP or any set of receiving SDPs in the construct. There is an implicit level of routing in this connectivity construct that is not present in the other connectivity constructs because the provider's network must determine to which receiving SDPs to deliver each packet. This construct may be used to support P2P traffic between any pair of SDPs, or to support multicast or broadcast traffic from one SDP to a set of other SDPs. In the latter case, whether the service is delivered using multicast within the provider's network or using "ingress replication" or some other means is out of scope of the specification of the service. A service provider may choose to support A2A constructs, but to limit the traffic to unicast.¶

  The SLOs/SLEs in an A2A connectivity construct apply to individual sending SDPs regardless of the receiving SDPs, and there is no linkage between sender and receiver in the specification of the connectivity construct. A sending SDP may be "disappointed" if the receiver is over-subscribed. If a customer wants to be more specific about different behaviors from one SDP to another SDP, they should use P2P connectivity constructs.¶

A customer traffic flow may be unicast or multicast, and various network realizations are possible:¶

• Unicast traffic may be mapped to a P2P connectivity construct for direct delivery, or to an A2A connectivity construct for the service provider to perform routing to the destination SDP. It would be unusual to use a P2MP connectivity construct to deliver unicast traffic because all receiving SDPs would get a copy, but this can still be done if the receivers are capable of dropping the unwanted traffic.¶
• A bidirectional unicast service can be constructed by specifying two P2P connectivity constructs. An additional SLE may specify fate-sharing in this case.¶
• Multicast traffic may be mapped to a set of P2P connectivity constructs, a single P2MP connectivity construct, or a mixture of P2P and P2MP connectivity constructs. Multicast may also be supported by an A2A connectivity construct. The choice clearly influences how and where traffic is replicated in the network. With a P2MP or A2A connectivity construct, it is the operator's choice whether to realize the construct with ingress replication, multicast in the core, P2MP tunnels, or hub-and-spoke. This choice should not change how the customer perceives the service.¶
• The concept of a multipoint-to-point (MP2P) service can be realized with multiple P2P connectivity constructs. Note that, in this case, the egress may simultaneously receive traffic from all ingresses. The SLOs at the sending SDPs must be set with this in mind because the provider's network is not capable of coordinating the policing of traffic across multiple distinct source SDPs. It is assumed that the customer, requesting SLOs for the various P2P connectivity constructs, is aware of the capabilities of the receiving SDP. If the receiver receives more traffic than it can handle, it may drop some and introduce queuing delays.¶
• The concept of a multipoint-to-multipoint (MP2MP) service can best be realized using a set of P2MP connectivity constructs, but could be delivered over an A2A connectivity construct if each sender is using multicast. As with MP2P, the customer is assumed to be familiar with the capabilities of all receivers. A customer may wish to achieve an MP2MP service using a hub-and-spoke architecture where they control the hub: that is, the hub may be an SDP or an ancillary SDP (see Section 3.2.1) and the service may be achieved by using a set of P2P connectivity constructs to the hub, and a single P2MP connectivity construct from the hub.¶

From the above, it can be seen that the SLOs of the senders define the SLOs for the receivers on any connectivity construct. That is, and in particular, the network may be expected to handle the traffic volume from a sender to all destinations. This extends to all connectivity constructs in an IETF Network Slice service.¶

Note that the realization of an IETF Network Slice service does not need to map the connectivity constructs one-to-one onto underlying network constructs (such as tunnels, etc.). The service provided to the customer is distinct from how the provider decides to deliver that service.¶

If a CE has multiple attachment circuits to a PE within a given IETF Network Slice service and they are operating in single-active mode, then all traffic between the CE and its attached PEs transits a single attachment circuit; if they are operating in all-active mode, then traffic between the CE and its attached PEs is distributed across all of the active attachment circuits.¶

A given sending SDP may be part of multiple connectivity constructs within a single IETF Network Slice service, and the SDP may have different SLOs and SLEs for each connectivity construct to which it is sending. Note that a given sending SDP's SLOs and SLEs for a given connectivity construct apply between it and each of the receiving SDPs for that connectivity construct.¶

An IETF Network Slice service provider may freely make a deployment choice as to whether to offer a 1:1 relationship between IETF Network Slice service and connectivity construct, or to support multiple connectivity constructs in a single IETF Network Slice service. In the former case, the provider might need to deliver multiple IETF Network Slice services to achieve the function of the second case.¶

It should be noted that per Section 9 of [RFC4364] an IETF Network Slice service customer may actually provide IETF Network Slice services to other customers in a mode sometimes referred to as "carrier's carrier". In this case, the underlying IETF Network Slice service provider may be owned and operated by the same or a different provider network. As noted in Section 4.3, network slices may be composed hierarchically or serially.¶

Section 4.2 provides a description of endpoints in the context of IETF network slicing. These are known as Service Demarcation Points (SDPs). For a given IETF Network Slice service, the customer and provider agree, on a per-SDP basis which end of the attachment circuit provides the SDP (i.e., whether the attachment circuit is inside or outside the IETF Network Slice service). This determines whether the attachment circuit is subject to the set of SLOs and SLEs at the specific SDP.¶

4.2. IETF Network Slice Service Demarcation Points

As noted in Section 3.1, an IETF Network Slice provides connectivity between sets of SDPs with specific SLOs and SLEs. Section 3.2 goes on to describe how the IETF Network Slice service is composed of a set of one or more connectivity constructs that describe connectivity between the Service Demarcation Points (SDPs) across the underlay network.¶

The characteristics of IETF Network Slice SDPs are as follows.¶

• SDPs are conceptual points of connection to an IETF Network Slice. As such, they serve as the IETF Network Slice ingress/egress points.¶
• Each SDP maps to a device, application, or a network function, such as (but not limited to) routers, switches, interfaces/ports, firewalls, WAN, 4G/5G RAN nodes, 4G/5G Core nodes, application accelerators, server load balancers, NAT44 [RFC3022], NAT64 [RFC6146], HTTP header enrichment functions, and Performance Enhancing Proxies (PEPs) [RFC3135].¶
• An SDP is identified by a unique identifier in the context of an IETF Network Slice customer.¶
• The provider associates each SDP with a set of provider-scope identifiers such as IP addresses, encapsulation-specific identifiers (e.g., VLAN tag, MPLS Label), interface/port numbers, node ID, etc.¶

• SDPs are mapped to endpoints of services/tunnels/paths within the IETF Network Slice during its initialization and realization.¶

  • A combination of the SDP identifier and SDP provider-network-scope identifiers define an SDP in the context of the Network Slice Controller (NSC) (see Section 5.3).¶
  • The NSC will use the SDP provider-network-scope identifiers as part of the process of realizing the IETF Network Slice.¶

For a given IETF Network Slice service, the IETF Network Slice customer and provider agree where the endpoint (i.e., the service demarcation point) is located. This determines what resources at the edge of the network form part of the IETF Network Slice and are subject to the set of SLOs and SLEs for a specific endpoint.¶

Figure 1 shows different potential scopes of an IETF Network Slice that are consistent with the different SDP locations. For the purpose of this discussion and without loss of generality, the figure shows customer edge (CE) and provider edge (PE) nodes connected by attachment circuits (ACs). Notes after the figure give some explanations.¶

         |<---------------------- (1) ---------------------->|
         |                                                   |
         | |<-------------------- (2) -------------------->| |
         | |                                               | |
         | |        |<----------- (3) ----------->|        | |
         | |        |                             |        | |
         | |        |  |<-------- (4) -------->|  |        | |
         | |        |  |                       |  |        | |
         V V   AC   V  V                       V  V   AC   V V
     +-----+   |    +-----+                 +-----+    |   +-----+
     |     |--------|     |                 |     |--------|     |
     | CE1 |   |    | PE1 |. . . . . . . . .| PE2 |    |   | CE2 |
     |     |--------|     |                 |     |--------|     |
     +-----+   |    +-----+                 +-----+    |   +-----+
        ^              ^                       ^              ^
        |              |                       |              |
        |              |                       |              |
     Customer       Provider                Provider       Customer
     Edge 1         Edge 1                  Edge 2         Edge 2

Explanatory notes for Figure 1 are as follows:¶

1. If the CE is operated by the IETF Network Slice service provider, then the edge of the IETF Network Slice may be within the CE. In this case the slicing process may utilize resources from within the CE such as buffers and queues on the outgoing interfaces.¶
2. The IETF Network Slice may be extended as far as the CE, to include the AC, but not to include any part of the CE. In this case, the CE may be operated by the customer or the provider. Slicing the resources on the AC may require the use of traffic tagging (such as through Ethernet VLAN tags) or may require traffic policing at the AC link ends.¶
3. In another model, the SDPs of the IETF Network Slice are the customer-facing ports on the PEs. This case can be managed in a way that is similar to a port-based VPN: each port (AC) or virtual port (e.g., VLAN tag) identifies the IETF Network Slice and maps to an IETF Network Slice SDP.¶
4. Finally, the SDP may be within the PE. In this mode, the PE classifies the traffic coming from the AC according to information (such as the source and destination IP addresses, payload protocol and port numbers, etc.) in order to place it onto an IETF Network Slice.¶

The choice of which of these options to apply is entirely up to the network operator. It may limit or enable the provisioning of particular managed services and the operator will want to consider how they want to manage CEs and what control they wish to offer the customer over AC resources.¶

Note that Figure 1 shows a symmetrical positioning of SDPs, but this decision can be taken on a per-SDP basis through agreement between the customer and provider.¶

In practice, it may be necessary to map traffic not only onto an IETF Network Slice, but also onto a specific connectivity construct if the IETF Network Slice supports more than one with a source at the specific SDP. The mechanism used will be one of the mechanisms described above, dependent on how the SDP is realized.¶

Finally, note (as described in Section 2.1) that an SDP is an abstract endpoint of an IETF Network Slice service and as such may be a device, interface, or software component and may, in the case of network functions virtualization (for example), be an abstract function supported within the provider's network.¶

4.3. IETF Network Slice Composition

Operationally, an IETF Network Slice may be composed of two or more IETF Network Slices as specified below. Decomposed network slices are independently realized and managed.¶

• Hierarchical (i.e., recursive) composition: An IETF Network Slice can be further sliced into other network slices. Recursive composition allows an IETF Network Slice at one layer to be used by the other layers. This type of multi-layer vertical IETF Network Slice associates resources at different layers.¶
• Sequential composition: Different IETF Network Slices can be placed into a sequence to provide an end-to-end service. In sequential composition, each IETF Network Slice would potentially support different dataplanes that need to be stitched together.¶

5.1. IETF Network Slice Stakeholders

An IETF Network Slice and its realization involves the following stakeholders. The IETF Network Slice customer and IETF Network Slice provider (see Section 2.1) are also stakeholders.¶

Orchestrator:

An orchestrator is an entity that composes different services, resource, and network requirements. It interfaces with the IETF NSC when composing a complex service such as an end-to-end network slice.¶

IETF Network Slice Controller (NSC):

The NSC realizes an IETF Network Slice in the underlay network, and maintains and monitors the run-time state of resources and topologies associated with it. A well-defined interface is needed to support interworking between different NSC implementations and different orchestrator implementations.¶

Network Controller:

The Network Controller is a form of network infrastructure controller that offers network resources to the NSC to realize a particular network slice. This may be an existing network controller associated with one or more specific technologies that may be adapted to the function of realizing IETF Network Slices in a network.¶

5.2. Expressing Connectivity Intents

An IETF Network Slice customer communicates with the NSC using the IETF Network Slice Service Interface.¶

An IETF Network Slice customer may be a network operator who, in turn, uses the IETF Network Slice to provide a service for another IETF Network Slice customer.¶

Using the IETF Network Slice Service Interface, a customer expresses requirements for a particular slice by specifying what is required rather than how that is to be achieved. That is, the customer's view of a slice is an abstract one. Customers normally have limited (or no) visibility into the provider network's actual topology and resource availability information.¶

This should be true even if both the customer and provider are associated with a single administrative domain, in order to reduce the potential for adverse interactions between IETF Network Slice customers and other users of the underlay network infrastructure.¶

The benefits of this model can include the following.¶

• Security: The underlay network components are less exposed to attack because the underlay network (or network operator) does not need to expose network details (topology, capacity, etc.) to the IETF Network Slice customers.¶
• Layered Implementation: The underlay network comprises network elements that belong to a different layer network than customer applications. Network information (advertisements, protocols, etc.) that a customer cannot interpret or respond to is not exposed to the customer. (Note - a customer should not use network information not exposed via the IETF Network Slice Service Interface, even if that information is available.)¶
• Scalability: Customers do not need to know any information concerning Network topology, capabilities, or state beyond that which is exposed via the IETF Network Slice Service Interface.¶

The general issues of abstraction in a TE network are described more fully in [RFC7926].¶

This framework document does not assume any particular technology layer at which IETF Network Slices operate. A number of layers (including virtual L2, Ethernet or, IP connectivity) could be employed.¶

Data models and interfaces are needed to set up IETF Network Slices, and specific interfaces may have capabilities that allow creation of slices within specific technology layers.¶

Layered virtual connections are comprehensively discussed in other IETF documents. See, for instance, GMPLS-based networks [RFC5212] and [RFC4397], or Abstraction and Control of TE Networks (ACTN) [RFC8453] and [RFC8454]. The principles and mechanisms associated with layered networking are applicable to IETF Network Slices.¶

There are several IETF-defined mechanisms for expressing the need for a desired logical network. The IETF Network Slice Service Interface carries data either in a protocol-defined format, or in a formalism associated with a modeling language.¶

For instance:¶

• The Path Computation Element (PCE) Communication Protocol (PCEP) [RFC5440] and GMPLS User-Network Interface (UNI) using RSVP-TE [RFC4208] use a TLV-based binary encoding to transmit data.¶
• The Network Configuration Protocol (NETCONF) [RFC6241] and RESTCONF Protocol [RFC8040] use XML and JSON encoding.¶
• gRPC/GNMI [I-D.openconfig-rtgwg-gnmi-spec] uses a binary encoded programmable interface. ProtoBufs can be used to model gRPC and GNMI data.¶
• For data modeling, YANG ([RFC6020] and [RFC7950]) may be used to model configuration and other data for NETCONF, RESTCONF, and GNMI, among others.¶

While several generic formats and data models for specific purposes exist, it is expected that IETF Network Slice management may require enhancement or augmentation of existing data models. Further, it is possible that mechanisms will be needed to determine the feasibility of service requests before they are actually made.¶

5.3. IETF Network Slice Controller (NSC)

The IETF NSC takes abstract requests for IETF Network Slices and implements them using a suitable underlay technology. An IETF NSC is the key component for control and management of the IETF Network Slice. It provides the creation/modification/deletion, monitoring and optimization of IETF Network Slices in a multi-domain, a multi-technology and multi-vendor environment.¶

The main task of the IETF NSC is to map abstract IETF Network Slice requirements to concrete technologies and establish required connectivity ensuring that resources are allocated to the IETF Network Slice as necessary.¶

The IETF Network Slice Service Interface is used for communicating details of an IETF Network Slice (configuration, selected policies, operational state, etc.), as well as information about status and performance of the IETF Network Slice. The details for this IETF Network Slice Service Interface are not in scope for this document.¶

The controller provides the following functions.¶

• Provides an IETF Network Slice Service Interface for creation/modification/deletion of the IETF Network Slices that is agnostic to the technology of the underlay network. The API exposed by this interface communicates the Service Demarcation Points of the IETF Network Slice, IETF Network Slice SLO/SLE parameters (and possibly monitoring thresholds), applicable input selection (filtering) and various policies, and provides a way to monitor the slice.¶
• Determines an abstract topology connecting the SDPs of the IETF Network Slice that meets criteria specified via the IETF Network Slice Service Interface. The NSC also retains information about the mapping of this abstract topology to underlay components of the IETF Network Slice as necessary to monitor IETF Network Slice status and performance.¶

• Provides "Mapping Functions" for the realization of IETF Network Slices. In other words, it will use the mapping functions that:¶

  • map IETF Network Slice Service Interface requests that are agnostic to the technology of the underlay network to technology-specific network configuration interfaces.¶
  • map filtering/selection information as necessary to entities in the underlay network so that those entities are able to identify what traffic is associated with which connectivity construct and IETF network slice and necessary according to the realization solution, and how traffic should be treated to meet the SLOs and SLEs of the connectivity construct.¶
• The controller collects telemetry data (e.g., OAM results, statistics, states, etc.) via a network configuration interface for all elements in the abstract topology used to realize the IETF Network Slice.¶
• Evaluates the current performance against IETF Network Slice SLO parameters using the telemetry data from the underlying realization of an IETF Network Slice (i.e., services/paths/tunnels). Exposes this performance to the IETF Network Slice customer via the IETF Network Slice Service Interface. The IETF Network Slice Service Interface may also include the capability to provide notifications if the IETF Network Slice performance reaches threshold values defined by the IETF Network Slice customer.¶

     +------------------------------------------+
     | Customer higher level operation system   |
     |   (e.g E2E network slice orchestrator,   |
     |      customer network management system) |
     +------------------------------------------+
                          A
                          | IETF Network Slice Service Interface
                          V
     +------------------------------------------+
     |    IETF Network Slice Controller (NSC)   |
     +------------------------------------------+
                          A
                          | Network Configuration Interface
                          V
     +------------------------------------------+
     |           Network Controllers            |
     +------------------------------------------+

      --------------
     | Network      |
     | Slice        |
     | Orchestrator |
      --------------
       | IETF Network Slice
       | Service Request
       |                       Customer view
   ....|................................
      -v-------------------    Operator view
     |Controller           |
     |  ------------       |
     | | IETF       |      |
     | | Network    |      |--> Virtual Network
     | | Slice      |      |
     | | Controller |      |
     | | (NSC)      |      |
     |  ------------       |
   ..|     | Network       |............
     |     | Configuration |   Underlay Network
     |     v               |
     |  ------------       |
     | | Network    |      |
     | | Controller |      |
     | | (NC)       |      |
     |  ------------       |
      ---------------------
       | Device Configuration
       v

6.1. Architecture to Realize IETF Network Slices

The architecture described in this section is deliberately at a high level. It is not intended to be prescriptive: implementations and technical solutions may vary freely. However, this approach provides a common framework that other documents may reference in order to facilitate a shared understanding of the work.¶

Figure 4 shows the architectural components of a network managed to provide IETF Network Slices. The customer's view is of individual IETF Network Slices with their SDPs, and connectivity constructs. Requests for IETF Network Slices are delivered to the NSC.¶

The figure shows, without loss of generality, the CEs, ACs, and PEs, that exist in the network. The SDPs are not shown and can be placed in any of the ways described in Section 4.2.¶

                       --      --      --
                      |CE|    |CE|    |CE|
                       --      --      --
                     AC :    AC :    AC :
                     ----------------------       -------
                    ( |PE|....|PE|....|PE| )     ( IETF  )
   IETF Network    (   --:     --     :--   )   ( Network )
   Slice Service   (     :............:     )   (  Slice  )
   Request          (  IETF Network Slice  )     (       )  Customer
     v               ----------------------       -------     View
     v        ............................\........./...............
     v                                     \       /        Provider
     v    >>>>>>>>>>>>>>>  Grouping/Mapping v     v           View
     v   ^             -----------------------------------------
     v   ^            ( |PE|.......|PE|........|PE|.......|PE|  )
    ---------        (   --:        --         :--         --    )
   |         |       (     :...................:                 )
   |   NSC   |        (        Network Resource Partition       )
   |         |         -----------------------------------------
   |         |                             ^
   |         |>>>>>  Resource Partitioning |
    ---------          of Filter Topology  |
     v   v                                 |
     v   v            -----------------------------      --------
     v   v           (|PE|..-..|PE|... ..|PE|..|PE|)    (        )
     v   v          ( :--  |P|  --   :-:  --   :--  )  (  Filter  )
     v   v          ( :.-   -:.......|P|       :-   )  ( Topology )
     v   v          (  |P|...........:-:.......|P|  )   (        )
     v   v           (  -    Filter Topology       )     --------
     v   v            -----------------------------       ^
     v    >>>>>>>>>>>>  Topology Filter ^                /
     v        ...........................\............../...........
     v                                    \            /  Underlay
    ----------                             \          /  (Physical)
   |          |                             \        /    Network
   | Network  |    ----------------------------------------------
   |Controller|   ( |PE|.....-.....|PE|......    |PE|.......|PE| )
   |          |  (   --     |P|     --      :-...:--     -..:--   )
    ----------  (    :       -:.............|P|.........|P|        )
        v       (    -......................:-:..-       -         )
         >>>>>>> (  |P|.........................|P|......:        )
     Program the  (  -                           -               )
       Network     ----------------------------------------------

The network itself (at the bottom of the figure) comprises an underlay network. This could be a physical network, but may be a virtual network. The underlay network is provisioned through network controllers that may utilize device controllers [RFC8309].¶

The underlay network may optionally be filtered or customized by the network operator to produce a number of network topologies that we call Filter Topologies. Customization is just a way of selecting specific resources (e.g., nodes and links) from the underlay network according to their capabilities and connectivity in the underlay network. These actions are configuration options or operator policies. The resulting topologies can be used as candidates to host IETF Network Slices and provide a useful way for the network operator to know in advance that all of the resources they are using to plan an IETF Network Slice would be able to meet specific SLOs and SLEs. The creation of a Filter Topology could be an offline planning activity or could be performed dynamically as new demands arise. The use of Filter Topologies is entirely optional in the architecture, and IETF Network Slices could be hosted directly on the underlay network.¶

Recall that an IETF Network Slice is a service requested by / provided for the customer. The IETF Network Slice service is expressed in terms of one or more connectivity constructs. An implementation or operator is free to limit the number of connectivity constructs in a slice to exactly one. Each connectivity construct is associated within the IETF Network Slice service request with a set of SLOs and SLEs. The set of SLOs and SLEs does not need to be the same for every connectivity construct in the slice, but an implementation or operator is free to require that all connectivity constructs in a slice have the same set of SLOs and SLEs.¶

One or more connectivity constructs from one or more slices are mapped to a set of network resources called a Network Resource Partition (NRP). A single connectivity construct is mapped to only one NRP (that is, the relationship is many to one). An NRP may be chosen to support a specific connectivity construct because of its ability to support a specific set of SLOs and SLEs, or its ability to support particular connectivity types, or for any administrative or operational reason. An implementation or operator is free to map each connectivity construct to a separate NRP, although there may be scaling implications depending on the solution implemented. Thus, the connectivity constructs from one slice may be mapped to one or more NRPs. By implication from the above, an implementation or operator is free to map all the connectivity constructs in a slice to a single NRP, and to not share that NRP with connectivity constructs from another slice.¶

An NRP is simply a collection of resources identified in the underlay network. Thus, the NRP is a scoped view of a topology and may be considered as a topology in its own right. The process of determining the NRP may be made easier if the underlay network topology is first filtered into a Filter Topology in order to be aware of the subset of network resources that are suitable for specific NRPs, but this is optional.¶

The steps described here can be applied in a variety of orders according to implementation and deployment preferences. Furthermore, the steps may be iterative so that the components are continually refined and modified as network conditions change and as service requests are received or relinquished, and even the underlay network could be extended if necessary to meet the customers' demands.¶

6.2. Procedures to Realize IETF Network Slices

There are a number of different technologies that can be used in the underlay, including physical connections, MPLS, time-sensitive networking (TSN), Flex-E, etc.¶

An IETF Network Slice can be realized in a network, using specific underlay technology or technologies. The creation of a new IETF Network Slice will be realized with following steps:¶

• The NSC exposes the network slicing capabilities that it offers for the network it manages so that the customer can determine whether to request services and what features are in scope.¶
• The customer may issue a request to determine whether a specific IETF Network Slice could be supported by the network. The NSC may respond indicating a simple yes or no, and may supplement a negative response with information about what it could support were the customer to change some requirements.¶
• The customer requests an IETF Network Slice. The NSC may respond that the slice has or has not been created, and may supplement a negative response with information about what it could support were the customer to change some requirements.¶
• When processing a customer request for an IETF Network Slice, the NSC maps the request to the network capabilities and applies provider policies before creating or supplementing the NRP.¶

Regardless of how IETF Network Slice is realized in the network (i.e., using tunnels of different types), the definition of the IETF Network Slice service does not change at all. The only difference is how the slice is realized. The following sections briefly introduce how some existing architectural approaches can be applied to realize IETF Network Slices.¶

6.3. Applicability of ACTN to IETF Network Slices

Abstraction and Control of TE Networks (ACTN - [RFC8453]) is a management architecture and toolkit used to create virtual networks (VNs) on top of a TE underlay network. The VNs can be presented to customers for them to operate as private networks.¶

In many ways, the function of ACTN is similar to IETF network slicing. Customer requests for connectivity-based overlay services are mapped to dedicated or shared resources in the underlay network in a way that meets customer guarantees for service level objectives and for separation from other customers' traffic. [RFC8453] describes the function of ACTN as collecting resources to establish a logically dedicated virtual network over one or more TE networks. Thus, in the case of a TE-enabled underlay network, the ACTN VN can be used as a basis to realize IETF network slicing.¶

While the ACTN framework is a generic VN framework that can be used for VN services beyond the IETF Network Slice, it also a suitable basis for delivering and realizing IETF Network Slices.¶

Further discussion of the applicability of ACTN to IETF Network Slices including a discussion of the relevant YANG models can be found in [I-D.ietf-teas-applicability-actn-slicing].¶

6.4. Applicability of Enhanced VPNs to IETF Network Slices

An enhanced VPN (VPN+) is designed to support the needs of new applications, particularly applications that are associated with 5G services, by utilizing an approach that is based on existing VPN and TE technologies and adds characteristics that specific services require over and above VPNs as they have previously been specified.¶

An enhanced VPN can be used to provide enhanced connectivity services between customer sites and can be used to create the infrastructure to underpin a IETF Network Slice service.¶

It is envisaged that enhanced VPNs will be delivered using a combination of existing, modified, and new networking technologies.¶

[I-D.ietf-teas-enhanced-vpn] describes the framework for Enhanced Virtual Private Network (VPN+) services.¶

6.5. Network Slicing and Aggregation in IP/MPLS Networks

Network slicing provides the ability to partition a physical network into multiple isolated logical networks of varying sizes, structures, and functions so that each slice can be dedicated to specific services or customers.¶

Many approaches are currently being worked on to support IETF Network Slices in IP and MPLS networks with or without the use of Segment Routing. Most of these approaches utilize a way of marking packets so that network nodes can apply specific routing and forwarding behaviors to packets that belong to different IETF Network Slices. Different mechanisms for marking packets have been proposed (including using MPLS labels and Segment Routing segment IDs) and those mechanisms are agnostic to the path control technology used within the underlay network.¶

These approaches are also sensitive to the scaling concerns of supporting a large number of IETF Network Slices within a single IP or MPLS network, and so offer ways to aggregate the connectivity constructs of slices (or whole slices) so that the packet markings indicate an aggregate or grouping where all of the packets are subject to the same routing and forwarding behavior.¶

At this stage, it is inappropriate to mention any of these proposed solutions that are currently work in progress and not yet adopted as IETF work.¶

6.6. Network Slicing and Service Function Chaining (SFC)

A customer may request an IETF Network Slice service that involves a set of service functions (SFs) together with the order in which these SFs are invoked. Also, the customer can specify the service objectives to be met by the underly network (e.g., one-way delay to cross a service function path, one-way delay to reach a specific SF). These SFs are considered as ancillary SDPs and are possibly placeholders (i.e., the SFs are identified, but not their locators).¶

Service Function Chaining (SFC) [RFC7665] techniques can be used by a provider to instantiate such an IETF Network Service Slice. The NSC may proceed as follows.¶

• Expose a set of ancillary SDPs that are hosted in the underlay network.¶
• Capture the SFC requirements (including, traffic performance metrics) from the customer. One or more service chains may be associated with the same IETF Network Slice service as connectivity constructs.¶
• Execute an SF placement algorithm to decide where to locate the ancillary SDPs in order to fulfil the service objectives.¶

• Generate SFC classification rules to identify (part of) the slice traffic that will be bound to an SFC. These classification rules may be the same as or distinct from the identification rules used to bind incoming traffic to the associated IETF Network Slice.¶

  The NSC also generates a set of SFC forwarding policies that govern how the traffic will be forwarded along a service function path (SFP).¶

• Identify the appropriate Classifiers in the underlay network and provision them with the classification rules. Likewise, the NSC communicates the SFC forwarding polices to the appropriate Service Function Forwarders (SFF).¶

The provider can enable an SFC data plane mechanism, such as [RFC8300], [RFC8596], or [I-D.ietf-spring-nsh-sr].¶

7.1. Isolation as a Service Requirement

An IETF Network Slice customer may request that the IETF Network Slice delivered to them is such that changes to other IETF Network Slices or to other services do not have any negative impact on the delivery of the IETF Network Slice. The IETF Network Slice customer may specify the degree to which their IETF Network Slice is unaffected by changes in the provider network or by the behavior of other IETF Network Slice customers. The customer may express this via an SLE it agrees with the provider. This concept is termed 'isolation'.¶

In general, a customer cannot tell whether a service provider is meeting an isolation SLE. If the service varies such that an SLO is breached then the customer will become aware of the problem, and if the service varies within the allowed bounds of the SLOs, there may be no noticeable indication that this SLE has been violated.¶

7.2. Isolation in IETF Network Slice Realization

Isolation may be achieved in the underlay network by various forms of resource partitioning ranging from dedicated allocation of resources for a specific IETF Network Slice, to sharing of resources with safeguards. For example, traffic separation between different IETF Network Slices may be achieved using VPN technologies, such as L3VPN, L2VPN, EVPN, etc. Interference avoidance may be achieved by network capacity planning, allocating dedicated network resources, traffic policing or shaping, prioritizing in using shared network resources, etc. Finally, service continuity may be ensured by reserving backup paths for critical traffic, dedicating specific network resources for a selected number of IETF Network Slices.¶