Internet DRAFT - draft-ali-spring-network-slicing-building-blocks
draft-ali-spring-network-slicing-building-blocks
SPRING Working Group Z. Ali
Internet-Draft C. Filsfils
Intended status: Standards Track P. Camarillo
Expires: August 21, 2021 Cisco Systems
D. Voyer
Bell Canada
S. Matsushima
Softbank
February 21, 2021
Building blocks for Slicing in Segment Routing Network
draft-ali-spring-network-slicing-building-blocks-04.txt
Abstract
This document describes how to build network slice using the
Segment Routing based technology. It explains how the
building blocks specified for the Segment Routing can be used for
this purpose.
Status of This Memo
This Internet-Draft is submitted in full conformance with the
provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering
Task Force (IETF). Note that other groups may also distribute
working documents as Internet-Drafts. The list of current Internet-
Drafts is at https://datatracker.ietf.org/drafts/current/.
Internet-Drafts are draft documents valid for a maximum of six months
and may be updated, replaced, or obsoleted by other documents at any
time. It is inappropriate to use Internet-Drafts as reference
material or to cite them other than as "work in progress."
This Internet-Draft will expire on August 21, 2021.
Copyright Notice
Copyright (c) 2021 IETF Trust and the persons identified as the
document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal
Provisions Relating to IETF Documents
(https://trustee.ietf.org/license-info) in effect on the date of
publication of this document. Please review these documents
carefully, as they describe your rights and restrictions with respect
to this document. 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.
[Page 1]
Internet-Draft Network Slicing Using SR
Table of Contents
1 Introduction...................................................2
2 Segment Routing Policy.........................................3
2.1 Flex-Algorithm Based SR Policies .........................4
2.2 On-demand SR policy ......................................5
2.3 Automatic Steering .......................................6
2.4 Inter-domain Considerations ..............................6
3 TI-LFA and Microloop Avoidance.................................7
4 SR VPN.........................................................7
5 Stateless Service Programming..................................7
6 Operations, Administration, and Maintenance (OAM)..............8
7 QoS............................................................8
8 Stateless Network Slice Identification.........................9
8.1 Stateless Slice Identification in SRv6....................9
8.2 Stateless Slice Identification in SR-MPLS................10
8 Orchestration at the Controller...............................10
9 Illustration..................................................10
10 Security Considerations ....................................10
11 IANA Considerations ........................................11
12 References .................................................11
12.1 Normative References ..................................11
7.2...........................................................11
13 Acknowledgments ............................................11
14 Contributors ...............................................11
1 Introduction
As more and more Service Providers and Enterprises operate a
single network infrastructure to support an ever-increasing
number of services, the ability to custom fit transport to
application needs is critically important. This includes
creating network slices with different characteristics can
coexist on top of the shared network infrastructure.
Network Slicing is meant to create (end-to-end) partitioned
network infrastructure that can be used to provide
differentiated connectivity behaviors to fulfill the
requirements of a diverse set of services. Services belonging to
different Network slices can be wholly disjoint or can share
different parts of the network infrastructure. Network Slicing
is one of the requirements in 5G [3GPP 23501].
Segment Routing enables Service Providers to support Network
Slicing without any additional protocol, other than the SR IGP
extensions. The network as a whole, in a distributed and
entirely automated manner, can share a single infrastructure
resource along multiple virtual services (slices). For example,
one slice is optimized continuously for low-cost transport; a
second Slice is optimized continuously for low-latency
transport; a third Slice is orchestrated to support disjoint
Internet-Draft Network Slicing Using SR
services, etc. The optimization objective of each of these
slices is programmable by the operator.
The Segment Routing specification already contains the various
building blocks required to create network slices. This includes
the following.
. SR Policy with or without Flexible Algorithm.
. TI-LFA with O(50 msec) protection in the slice underlay.
. SR VPN.
. SR Service Programming (NFV, SFC).
. Operation, Administration and Management (OAM) and
Performance Management (PM).
. QoS using DiffServ.
. Stateless Network Slice Identification
. Orchestration at the Controller.
Each of these building blocks works independently of each other.
Their functionality can be combined to satisfy service
provider's requirement for the Network Slicing. An external
controller plays an important role to orchestrate these building
blocks into a Slicing service.
This document elaborates on the attributes of each of these
building blocks for Network Slicing. The document also highlights
how services in each Slice can benefit from traffic engineering,
network function virtualization/ service chaining (service
programming), OAM, performance management, SDN readiness, O (50 msec)
TI-LFA protection, etc. features of SR while respecting resource
partitioning employed over the common networking infrastructure.
The document equally applicable to the MPLS and SRv6
instantiations of segment routing.
The following subsection elaborates on each of these build
blocks.
2 Segment Routing Policy
Segment Routing (SR) allows a headend node to steer a packet
flow along any path without creating intermediate per-flow
states [I-D.ietf-spring-segment-routing-policy]. The headend
node steers a flow into a Segment Routing Policy (SR Policy).
I.e., the SR Policy can be used to steer traffic along any
arbitrary path in the network. This allows operators to enforce
low-latency and / or disjoint paths, regardless of the normal
forwarding paths.
Internet-Draft Network Slicing Using SR
The SR policy is able to support various optimization objectives
[I-D.draft-filsfils-spring-sr-policy-considerations]. The
optimization objectives can be instantiated for the IGP metric
([RFC1195] [RFC2328] [RFC5340]) xor the TE metric ([RFC5305],
[RFC3630]) xor the latency extended TE metric ([RFC7810]
[RFC7471]). In addition, an SR policy is able to various
constraints, including inclusion and/or exclusion of TE
affinity, inclusion and/or exclusion of IP address, inclusion
and/or exclusion of SRLG, inclusion and/or exclusion of admin-
tag, maximum accumulated metric (IGP, TE, and latency), maximum
number of SIDs in the solution SID-List, maximum number of
weighted SID-Lists in the solution set, diversity to another
service instance (e.g., link, node, or SRLG disjoint paths
originating from different head-ends), etc. [I-D.draft-filsfils-
spring-sr-policy-considerations]. The supports for various
optimization objectives and constraints enables SR policy to
create Slices in the network.
SR policy can be instantiated with or without IGP Flexible
Algorithm feature. The following subsection describes the SR
Flexible Algorithm feature and how SR policy can utilize this
feature.
2.1 Flex-Algorithm
Flexible Algorithm enriches the SR Policy solution by adding
additional segments having different properties than the IGP
Prefix segments. Flex Algo adds flexible, user-defined segments
to the SRTE toolbox. Specifically, it allows for association of
the "intent" to Prefix SIDs. [I-D.ietf-lsr-flex-algo] defines
the IGP based Flex-Algorithm
solution which allows IGPs themselves to compute paths
constraint by the "intent" represented by the Flex-Algorithm.
The Flex-Algorithm has the following attributes:
. Algorithm associate to the SID a specific TE intent
expressed as an optimization objective (an algorithm) [I-
D.ietf-lsr-flex-algo].
. Flexibility includes the ability of network operators to
define the intent of each algorithm they implement.
. By design the mapping between the Flex-Algorithm and its
meaning is flexible and is defined by the user.
. Flexibility also includes ability for operators to make the
decision to exclude some specific links from the shortest
path computation, e.g.,
Internet-Draft Network Slicing Using SR
o operator 1 may define Algo 128 to compute the shortest
path for TE metric and exclude red affinity links.
o operator 2 may define Algo 128 to compute the shortest
path for latency metric and exclude blue affinity
links.
A Network Slice can be created by associating a Flexible-
Algorithm value with the Slice via provisioning.
Flex Alg leverages SR on-demand next hop (ODN) and Automated
Steering for intent-based instantiation of traffic engineered
paths described in the following sub-sections. Specifically, as
specified in [RFC8402] the IGP Flex Algo
Prefix SIDs can also be used as segments within SR Policies
thereby leveraging the underlying IGP Flex Algo solution.
2.2 On-demand SR policy
Segment Routing On-Demand Next-hop (ODN) functionality enables
on-demand creation of SR Policies for service traffic. Using a
Path Computation Element (PCE), end-to-end SR Policy paths can
be computed to provide end-to-end Segment Routing connectivity,
even in multi-domain networks running with or without IGP
Flexible-Algorithm [I-D.draft-ietf-spring-segment-routing-
policy].
The On-Demand Next-hop functionality provides optimized service
paths to meet customer and application SLAs (such as latency,
disjointness) without any pre-configured TE tunnel and with the
automatic steering of the service traffic on the SR Policy
without a static route, autoroute-announce, or policy-based
routing.
With this functionality, a Network Service Orchestrator can
deploy the service based on their requirements. The service
head-end router requests the PCE to compute the path for the
service and then instantiates an SR Policy with the computed
path and steers the service traffic into that SR Policy. If the
topology changes, the stateful PCE updates the SR Policy path.
This happens seamlessly, while TI-LFA protects the traffic in
case the topology change happened due to a failure.
Internet-Draft Network Slicing Using SR
2.3 Automatic Steering
Automatically steering traffic into a Network Slice is one of
the fundamental requirement for Slicing. That is made possible
by the "Automated Steering" functionality of SR. Specifically,
SR policy can be used for traffic engineer paths
within a slice, "automatically steer" traffic to the right slice
and connect IGP Flex-Algorithm domains sharing the same
"intent".
A headend can steer a packet flow into a valid SR Policy within
a slice in various ways [I-D.draft-ietf-spring-segment-routing-
policy]:
. Incoming packets have an active SID matching a local
Binding SID (BSID) at the headend.
. Per-destination Steering: incoming packets match a
BGP/Service route which recurses on an SR policy.
. Per-flow Steering: incoming packets match or recurse on a
forwarding array of where some of the entries are SR
Policies.
. Policy-based Steering: incoming packets match a routing
policy which directs them on an SR policy.
2.4 Inter-domain Considerations
The network slicing needs to be extended across multiple domains
such that each domain can satisfy the intent consistently. SR
has native inter-domain mechanisms, e.g., SR policies are
designed to span multiple domains using a PCE based solution [I-
D.ietf-spring-segment-routing], [I-D.ietf-spring-segment-
routing-central-epe]. An edge router upon service configuration
automatically requests to the Segment Routing PCE an inter-
domain path to the remote service endpoint. The path can either
be for simple best-effort inter-domain reachability or for
reachability with an SLA contract and can be restricted to a
Network Slice.
The SR native mechanisms for inter-domain are easily extendable
to include the case when different IGP Flex-Algorithm values are
used to represent the same intent. E.g., in domain1 Service
Provider 1 (SP1) may use flex-algo 128 to indicate low latency
Slice and in domain2 Service Provider 2 (SP2) may use flex-algo
129 to indicate low latency Slice. When an automation system at
a PE1 in SP1 network configures a service with next hop (PE2) in
SP2 network, SP1 contacts a Path Computation Element (PCE) to
find a route to PE2. In the request, the PE1 also indicates the
intent (i.e., the Flex-Algo 128) in the PCEP message. As the PCE
has a complete understanding of both Domains, it can understand
the path computation in Domain1 needs to be performed for
Algorithm 128 and path computation in Domain2 needs to be
Internet-Draft Network Slicing Using SR
performed for Algorithm 129 (i.e., in the Low Latency Network
Slice in both domains).
3 TI-LFA and Microloop Avoidance
The Segment Routing-based fast-reroute solution, TI-LFA, can
provide per-destination sub-50msec protection upon any single
link, node or SRLG failure regardless of the topology. The
traffic is rerouted straight to the post-convergence path, hence
avoiding any intermediate flap via an intermediate path. The
primary and backup path computation is completely automatic by
the IGP.
[I-D.draft-bashandy-rtgwg-segment-routing-ti-lfa] proposes a
Topology Independent Loop-free Alternate Fast Re-route (TI-LFA),
aimed at protecting node and adjacency segments within O(50
msec) in the Segment Routing networks. Furthermore, [I-D. draft-
bashandy-rtgwg-segment-routing-uloop] provides a mechanism
leveraging Segment Routing to ensure loop-freeness during the
IGP reconvergence process following a link-state change event.
As mentioned earlier, Network Slicing in Segment Routing works
seamlessly with all the other components of the Segment Routing.
This, of course, includes TI-LFA and microloop avoidance within
a Slice, with the added benefit that backup path only uses
resources available to the Slice. For example, when Flexible
Algorithm is used, the TI-LFA backup path computation is
performed such that it is optimized per Flexible-Algorithm. The
backup path shares the same properties are the primary path. The
backup path does not use a resource outside the Slice of the
primary path it is protecting.
4 SR VPN
Virtual Private Networks (VPNs) provides a mean for creating a
logically separated network to a different set of users access
to a common network. Segment Routing is equipped with the rich
multi-service virtual private network (VPN) capabilities,
including Layer 3 VPN (L3VPN), Virtual Private Wire Service
(VPWS), Virtual Private LAN Service (VPLS), and Ethernet VPN
(EVPN). The ability of Segment Routing to support different VPN
technologies is one of the fundamental building blocks for
creating slicing an SR network.
Internet-Draft Network Slicing Using SR
5 Stateless Service Programming
An important part of the Network Slicing is the orchestration of
virtualized service containers. [I-D.draft-xuclad-spring-sr-
service-chaining] describes how to implement service segments
and achieve stateless service programming in SR-MPLS and SRv6
networks. It introduces the notion of service segments. The
ability of encoding the service segments along with the
topological segment enables service providers to forward packets
along a specific network path, but also steer them through VNFs
or physical service appliances available in the network.
In an SR network, each of the service, running either on a
physical appliance or in a virtual environment, is associated
with a segment identifier (SID) for the service. These service
SIDs are then leveraged as part of a SID-list to steer packets
through the corresponding services. Service SIDs may be
combined with topological SIDs to achieve service programming
while steering the traffic through a specific topological path
in the network. In this fashion, SR provides a fully integrated
solution for overlay, underlay and service programming building
blocks needed to satisfy network slicing requirements.
6 Operations, Administration, and Maintenance (OAM)
There are various OAM elements that are critical to satisfy
Network Slicing requirements. These includes but not limited to
the following:
. Measuring per-link TE Matric.
. Flooding per-link TE Matric.
. Taking TE Matric into account during path calculation.
. Taking TE Matric bound into account during path calculation.
. SLA Monitoring: Service Provider can monitor each SR Policy
in a Slice to Monitor SLA offered by the Policy using
technique described in [I-D.draft-gandhi-spring-udp-pm].
This includes monitoring end-to-end delays on all ECMP
paths of the Policy as well as monitoring traffic loss on a
Policy. Remedial mechanisms can be used to ensure that the
SR policy conforms to the SLA contract.
7 QoS
Segment Routing relies on MPLS and IP Differentiated Services.
Differentiated services enhancements are intended to enable
scalable service discrimination in the Internet without the need
for per-flow state and signaling at every hop. [RFC2475] defines
Internet-Draft Network Slicing Using SR
an architecture for implementing scalable service
differentiation in the Internet. This architecture is composed
of many functional elements implemented in network nodes,
including a small set of per-hop forwarding behaviors, packet
classification functions, and traffic conditioning functions
including metering, marking, shaping, and policing.
The DiffServ architecture achieves scalability by implementing
complex classification and conditioning functions only at
network boundary nodes, and by applying per-hop behaviors to
aggregates of traffic depending on the traffic marker.
Specifically, the node at the ingress of the DiffServ domain
conditions, classifies and marks the traffic into a limited
number of traffic classes. The function is used to ensure that
the slice's traffic conforms to the contract associated with the
slice.
Per-hop behaviors are defined to permit a reasonably granular
means of allocating buffer and bandwidth resources at each node
among competing traffic streams. Specifically, per class
scheduling and queuing control mechanisms are applied at each IP
hop to the traffic classes depending on packet's marking.
Techniques such as queue management and a variety of scheduling
mechanisms are used to get the required packet behavior to meet
the slice's SLA.
8 Stateless Network Slice Identification
Some use-cases require a slice identifier (SLID) in the packet
to provide differentiated treatment of the packets belonging
to different network slices.
The network slice instantiation using the SLID in the packet
is required to work with the building blocks described in the
previous sections. For example, the QoS/ DiffServ needs to be
observed on a per slice basis. The slice identification needs
to be topologically independent and stateless.
8.1 Stateless Slice Identification in SRv6
[I-D.draft-filsfils-spring-srv6-stateless-slice-id] describes
a stateless encoding of slice identification in the outer IPv6
the header of an SRv6 domain. As defined in RFC8754 [RFC8754],
when an ingress PE receives a packet that traverses the SR domain,
it encapsulates the packet in an outer IPv6 header and an optional
SRH. Based on a local policy of the SR domain, the Flow Label field
of the outer IPv6 header carries the SLID. Specifically, the SLID
is added in the 8 most significant bits of the Flow Label field
of the outer IPv6 header. The remaining 12 bits of the Flow Label
field are set as described in section 5.5 of [RFC8754] for
inter-domain packets. Based on the local policy of the SR domain,
the draft also uses one of the bits in the Traffic Class field of
the outer IPv6 header to indicate that the entropy label contains
the SLID.
The network slicing mechanism described in
[I-D.draft-filsfils-spring-srv6-stateless-slice-id] works seamlessly
with the building blocks described in the previous sections. For
example, the slice identification is independent of topology and
the network's QoS/DiffServ policy. It enables scalable network
slicing for SRv6 overlays.
8.2 Stateless Slice Identification in SR-MPLS
[I-D.draft-decraene-mpls-slid-encoded-entropy-label-id] describes a
similar stateless encoding of slice identification in the SR-MPLS domain.
Specifically, the document extends the use of the Entropy Label to carry
the SLID. The number of bits to be used for encoding the SLID in the
Entropy Label is governed by a local policy of the SR domain. Based on
the local policy of the SR domain, the draft uses one of the bits in the
TTL field of the Entropy Label to indicate that the Entropy Label contains
the SLID.
The network slicing mechanism described in
[I-D.draft-decraene-mpls-slid-encoded-entropy-label-id] works seamlessly
with the building blocks described in the previous sections. For
example, the slice identification is independent of topology and
the network's QoS/DiffServ policy. It enables scalable network
slicing for SR-MPLS overlays.
Internet-Draft Network Slicing Using SR
8 Orchestration at the Controller
A controller plays a vital role in orchestrating the SR building
blocks discussed above to create Network Slices. The controller also
performs admission control and traffic placement for slice
management at the transport layer. The SDN friendliness of the
SR technology becomes handy to realize the orchestration. The
controller may use PCEP or Netconf to interact with the routers.
The router implements Yang model for SR-based network slicing.
Specification of the controller technology for orchestrating
Network Slices, services and admission control for the services
is outside the scope of this draft.
9 Illustration
To be added in a later revision.
10 Security Considerations
This document does not impose any additional security
challenges.
Internet-Draft Network Slicing Using SR
11 IANA Considerations
This document does not define any new protocol or any extension
to an existing protocol.
12 References
12.1 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,
<https://www.rfc-editor.org/info/rfc2119>.
7.2. Informative References
[I-D.ietf-spring-segment-routing-policy]
Filsfils, C., Sivabalan, et al, "Segment Routing Policy
For Traffic Engineering", draft-ietf-spring-segment-
routing-policy (work in progress).
[I-D.ietf-lsr-flex-algo] P. Psenak, et al,
draft-ietf-lsr-flex-algo, work in progress.
[RFC8402]
Filsfils, C., Previdi, S., Ginsberg, L., Decraene, B.,
Litkowski, S., and R. Shakir, "Segment Routing
Architecture", RFC8402.
[I-D.draft-filsfils-spring-sr-policy-considerations]
Filsfils, C., et al. draft-filsfils-spring-sr-policy-
considerations (work in progress)
[RFC8754]
Filsfils, C., Previdi, S., Leddy, J., Matsushima, S., and
d. daniel.voyer@bell.ca, "IPv6 Segment Routing Header
(SRH)", draft-ietf-6man-segment-routing-header-16 (work in
progress), February 2019.
[I-D.draft-filsfils-spring-srv6-stateless-slice-id] Filsfils, C., et al.
draft-filsfils-spring-srv6-stateless-slice-id, work in progress.
[I-D.draft-decraene-mpls-slid-encoded-entropy-label-id] Decraene B.,
Filsfils, C., Henderickx W., Saad T., Beeram V.,
work in progress.
13 Acknowledgments
14 Contributors
Francois Clad
Cisco Systems, Inc.
fclad@cisco.com
Internet-Draft Network Slicing Using SR
Authors' Addresses
Zafar Ali
Cisco Systems, Inc.
Email: zali@cisco.com
Clarence Filsfils
Cisco Systems, Inc.
Email: cf@cisco.com
Pablo Camarillo Garvia
Cisco Systems, Inc.
Email: pcamaril@cisco.com
Daniel Voyer
Bell Canada
Email: daniel.voyer@bell.ca
