RTG Working Group R. Raszuk, Ed.
Internet-Draft Bloomberg LP
Intended status: Informational September 26, 2019
Expires: March 29, 2020
IP Traffic Engineering Architecture with Network Programming
draft-raszuk-rtgwg-ip-te-np-00
Abstract
This document describes a control plane based IP Traffic Engineering
Architecture where path information is kept in the control plane by
selected nodes instead of being inserted into each packet on ingress
of an administrative domain. The described proposal is also fully
compatible with the concept of network programming.
It is positioned as a complimentary technique to native SRv6 and can
be used when there are concerns with increased packet size due to
depth of SID stack, possible concerns regarding exceeding MTU or more
strict simplicity requirements typically seen in number of enterprise
networks. The proposed solution is applicable to both IPv4 or IPv6
based networks.
As an additional added value, detection of end to end path liveness
as well as dynamic path selection based on real time path quality is
integrated from day one in the design.
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 March 29, 2020.
Raszuk Expires March 29, 2020 [Page 1]
Internet-Draft IP TE+NP September 2019
Copyright Notice
Copyright (c) 2019 IETF Trust and the persons identified as the
document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal
Provisions Relating to IETF Documents
(https://trustee.ietf.org/license-info) in effect on the date of
publication of this document. Please review these documents
carefully, as they describe your rights and restrictions with respect
to this document. Code Components extracted from this document must
include Simplified BSD License text as described in Section 4.e of
the Trust Legal Provisions and are provided without warranty as
described in the Simplified BSD License.
Table of Contents
1. Background . . . . . . . . . . . . . . . . . . . . . . . . . 2
2. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 3
3. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 4
4. Functional Description . . . . . . . . . . . . . . . . . . . 7
5. Control plane . . . . . . . . . . . . . . . . . . . . . . . . 10
6. Data plane . . . . . . . . . . . . . . . . . . . . . . . . . 11
7. Network Programming . . . . . . . . . . . . . . . . . . . . . 13
8. Active Path Probing . . . . . . . . . . . . . . . . . . . . . 15
8.1. TI-LFA Local Protection . . . . . . . . . . . . . . . . . 17
9. Solution advantages . . . . . . . . . . . . . . . . . . . . . 17
10. OAM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
11. Deployment considerations . . . . . . . . . . . . . . . . . . 19
12. Security considerations . . . . . . . . . . . . . . . . . . . 19
13. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 19
14. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 19
15. References . . . . . . . . . . . . . . . . . . . . . . . . . 19
15.1. Normative References . . . . . . . . . . . . . . . . . . 19
15.2. Informative References . . . . . . . . . . . . . . . . . 20
Author's Address . . . . . . . . . . . . . . . . . . . . . . . . 22
1. Background
Ability to steer data over selected topological points often
different from default IGP or BGP paths proves to provide substantial
advantages to consumers of such data. The construction of controlled
transit paths usually is driven by requirements to: offload
excessively used default routing paths, construct disjointed paths
for live-live dual streaming or create intra or inter-domain data
distribution overlays using dynamic real time SLAs criteria often
used along with per specific application mapping schema.
Raszuk Expires March 29, 2020 [Page 2]
Internet-Draft IP TE+NP September 2019
In addition to pure topological reasons there are often also
requirements for special data flow processing to happen in selected
network elements which by default would not be in the data path of
the subject flows. Examples of this could be: firewall traffic
screening, service function chaining, caching, deep packet
inspection, etc ...
While there are some solutions available to allow traffic engineering
in domains fully operated by single administrative entity there seems
to be lack of proposals which could be used to control
interconnections of sites over third party networks or Internet. As
part of that category one could also list public cloud tenancies
where ability to steer in/out traffic over other then default
Internet routing could provide much better SLA characteristics or
address some of the non purely technical requirements.
Another category of global networking which can significantly benefit
from standards based IP TE solution is unified model of path
engineering for Software Defined Wide Area Networks (SDWANs). One of
the basic operational principles in selected SDWANs is point to point
underlay selection based on the applied SLA characteristics. Adding
ability to traffic engineer such underlay flows allows to bypass
under performing underlay default paths or congestion points
occurring even few autonomous systems away.
2. Terminology
The following abbreviations are used within this document:
o TE - Traffic Engineering
o AF - Address Family
o IPv4 - Internet Protocol version 4
o IPv6 - Internet Protocol version 6
o IGP - Interior Gateway Protocol
o EH - Extension Header
o RIR - Regional Internet Registry
o PCE - Path Computation Element
o UDP - User Datagram Protocol
o BGP - Border Gateway Protocol
Raszuk Expires March 29, 2020 [Page 3]
Internet-Draft IP TE+NP September 2019
o SRH - Segment Routing Header
o OWAMP - A One-way Active Measurement Protocol
o DOH - Destination Option Header
o PE - Provider Edge
o SE - Segment Endpoint
o SID - Segment Identifier (PREFIX+FUNCTION+4bits}
o NMS - Network Management System
o CoS - Class of Service
o PCE - Path Computation Element
o PCEP - Path Computation Element Communication Protocol
o SR-MPLS - Segment Routing with MPLS data plane
o SRv6 - SRv6 Network Programming
o RTT - Round Trip Time
o MTU - Maximum Transmission Unit
o MOS - Mean Opinion Score
o OAM - Operation, Administration, Maintenance
o MPLS - Multiprotocol Label Switching
o GID - Group Identifier
3. Introduction
Proposed architecture described in this specification defines a new
forwarding paradigm which allows to create traffic engineered paths
either centrally or in a distributed way. With the assistance of
local provisioning tools or control plane such ordered set of paths
are distributed to those network elements which will participate in
data forwarding. In addition to basic packet forwarding the
architecture also provides mechanism to execution arbitrary
instructions at selected by operator network nodes which can include:
routers, switches, firewalls, service processors, hosts etc ...
Raszuk Expires March 29, 2020 [Page 4]
Internet-Draft IP TE+NP September 2019
Authors have taken a clean slate approach to look at the possible
options to engineer traffic within given administrative domain
boundaries. The solution is applicable to both traditional
"underlay" networks as well as administrative domains constructed
with "overlays". It is also 100% transparent to operating network
elements which would not participate in the traffic engineering
solution while maintaining packet's entropy and fast connectivity
restoration needs.
The proposed solution is constructed using either building blocks or
ideas borrowed from the following technologies:
o Segment Routing Architecture [RFC8402]
o Destination/Source Routing [I-D.ietf-rtgwg-dst-src-routing]
o Generic Packet Tunneling in IPv6 Specification [RFC2473]
o IP Encapsulation within IP [RFC2003]
o Encapsulating IP in UDP [I-D.xu-intarea-ip-in-udp]
o Advertising Segment Routing Policies in BGP
[I-D.ietf-idr-segment-routing-te-policy]
o BGP Vector Routing [I-D.patel-raszuk-bgp-vector-routing]
o A Path Computation Element (PCE) Based Architecture [RFC4655]
o PCEP Extensions for Segment Routing [I-D.ietf-pce-segment-routing]
o Topology Independent Fast Reroute using Segment Routing
[I-D.ietf-rtgwg-segment-routing-ti-lfa]
o A One-way Active Measurement Protocol (OWAMP) [RFC4656]
It is also fully compatible with following specifications to embed
network programming concept as is define in the below documents while
in the same time provides a new alternate encoding model:
o Internet Protocol, Version 6 (IPv6) Specification [RFC8200]
o IPv6 Segment Routing Header (SRH)
[I-D.ietf-6man-segment-routing-header]
o IPv4 Extension Headers and Flow Label [I-D.herbert-ipv4-eh]
Raszuk Expires March 29, 2020 [Page 5]
Internet-Draft IP TE+NP September 2019
o IPv4 Extension Headers and UDP Encapsulated Extension Headers
[I-D.herbert-ipv4-udpencap-eh]
For the intradomain Traffic Engineering needs the introduced overhead
is of fixed size and regardless of the amount of segment endpoints or
links which need to be traversed as part of the engineered path is
constant and equal to 28 octets for IPv4 and 40 octets for IPv6. If
additional segment end or path end instructions are to be added into
additional headers an extension header size will need to be included.
Instructions however, can also be embedded into SID destination or
reside above the encapsulation header. In those cases,3 the total
length of the overhead remains fixed as stated above.
Interdomain Traffic Engineering depending on the deployment model
could result in additional fixed 12 octets of the overhead. Overlay
deployment models will be discussed in more details in below Data
Plane section.
While the described architecture is applicable to both IPv4 and IPv6
networks the proposal could be split into separate documents each
focusing on specifics corresponding only to a single address family
if the community expresses such preference. However, due to the
number of common AF agnostic characteristics it is advised to keep it
within a single document.
Since the support of EH in IPv4 is planned to be introduced with a
rather limited scope, the end segment or end path instructions could
end up using other extension header types (for example: Destination
Options) in IPv4 packets or could be encoded into the destination
addresses itself. It has to be noted that IPv4 packets could be
encapsulated in IPv6 when carried across a given domain. The
document describes how the concept of network programming can be
applied without use of extension headers.
The proposal does not enforce any new dependencies on IP address
block allocations and is in full alignment to the current IETF and
RIRs address structure and allocation policies.
The core of the defined functionality does not require any new
protocol extensions. The solution attempts to maximize and reuse
extensions already defined. If more optimal protocol solutions
applicable to any of the defined functional blocks surface additional
work will take place in corresponding area/wg.
Described architecture does not belong to segment routing family even
if some terminology used to describe the proposal have been borrowed
from it. Major difference is that by design it uses control plane or
management plane to install per path state in the transit nodes
Raszuk Expires March 29, 2020 [Page 6]
Internet-Draft IP TE+NP September 2019
participating in the engineering of data paths instead of encoding
set of TE midpoints into each packet on ingress.
While scaling aspects of any solution is a very important factor it
needs to be put in perspective to the operational requirements as
well as characteristics of the designs. It also needs to be noted
that even basic IP routing is based on state in the network elements
and scale of Internet routing is usually orders of magnitude higher
then state of most traffic engineering needs. While looking at
scaling factors of the complete solution variable size per packet
overhead needs to be weighted against cost of additional per path
fixed size state in control and data plane.
IP TE+NP design while allowing operator to create centrally computed
and distribute strict end to end paths in number of deployments can
be used in fully distributed mode. Traffic steering decisions can
autonomously take place in any TE midpoint what is particularly
useful with all SLA or performance based routing deployments.
If there is any comparison to be made between SR and IP TE+NP
architectures putting aside other fundamental differences would be
the assumption of constructing segment routing paths only by Binding
SIDs (divided into static and variable parts) and only encoding them
at each segment endpoint in least significant bits of source and
destination address of the outer IP header.
4. Functional Description
For the purpose of this document the following term definitions will
be used in capital letter notation:
o CLASSIFIER_ID: Identifier to set of rules used for mapping flows
to TE paths. Length - 4 octets.
o PATH_GID_PFX: routable node prefix + locally significant PATH_GID
value. Length - 4 or 16 octets.
o SID: routable node prefix + opt. function + opt. parameters + 4
bits (Lookup Type) - Length - 4 or 16 octets.
o PATH_LIST: ordered list of SIDs. Length N x 4 or N x 16 octets.
N min = 1.
Raszuk Expires March 29, 2020 [Page 7]
Internet-Draft IP TE+NP September 2019
v------- ADMINISTRATIVE DOMAIN --------v
+---------------SE1--------------+
| |
| |
SRC_NET-----PE1----P1----SE2----P2----P3----SE3----PE2----DST_NET
| || |
| || |
+------- SE4 ------++----SE5----+
Basic Network Topology
Figure 1
Consider basic two requirements to be applied for some class of
transit traffic T1 and T2:
o T1: PATH_A1: PE1--SE1--PE2
o T2: PATH_A2: PE1--SE4--SE5--PE2
IGP metric of all interfaces is set to 10 except interfaces attached
to SE1, SE4 and SE5 nodes which are of metric of 100.
The shortest default path, in the example above, between PEs is:
PE1--P1--SE2--P2--P3--SE3--PE2
In order to accomplish the stated requirements (for traffic classes
T1 and T2 defined above) the following ordered path lists are created
in the control plane and either locally configured on both ingress
and segment endpoints or distributed by any of the control plane
protocols discussed in subsequent sections:
CLASSIFIER_ID: T1 CLASSIFIER_ID: T2
PATH_GID: A1 PATH_GID: A2
PATH_LIST: SE1, PE2 PATH_LIST: SE4, SE5, PE2
There are few core elements of the design as listed below:
o Each PATH_GID_PFX contains unique routable IP prefix from one of
the loopbacks of the corresponding ingress PE followed by PATH_GID
value (PATH GROUP-ID). For example, if the loopback's prefix is a
/64 IPv6 prefix there can be 2^64 unique paths originated at a
given PE. If the loopback address is a /16 IPv4 prefix (for
example used from [RFC1918] space) there can be 2^16 paths
initiating at a given IPv4 PE. The choice of mapping scheme is
local to the ingress PE and is assigned by the operator. Let's
observe that in most cases to describe reachability to the
Raszuk Expires March 29, 2020 [Page 8]
Internet-Draft IP TE+NP September 2019
PATH_GID_PFX only a single IGP loopback prefix may need to be
advertised from any ingress PE. It is also highly recommended
that such loopback prefixes configured on all ingress nodes
(ingress PEs) to be sourced from the same address block such that
it can be described by single aggregate prefix.
o Each PATH_LIST consists of a number of SID elements. Each SID is
a unique routable IP address from one of the loopbacks of the
corresponding Segment Endpoint (SE) node. For example, if the
loopback's prefix is a /64 IPv6 prefix there can be 2^(64-4)
unique SID terminating on a given node. If the loopback address
is a /16 IPv4 prefix (for example used from [RFC1918] space) there
can be 2^(16-4) SIDs present on a given IPv4 node. As defined, a
SID may represent not only a node's topological location in the
network (via IP prefix reachability), but it may also, optionally,
contain embedded functions with their parameters. In order to
even further help the forwarding layer within a given domain, the
last four bits can be consistently chosen to describe the lookup
type required to correctly switch a given packet.
o Upon ingress to the domain, and after classification, packets are
encapsulated into an additional outer IP header with the following
elements corresponding to the non-default forwarding requirements:
Classified as T1 flows: Classified as T2 flows:
----------------------- -----------------------
Source address: PATH_A1_PFX Source address: PATH_A2_PFX
Destination address: SID_SE1 Destination address: SID_SE4
In the case of IPv6 the encapsulation for the basic TE only
requirement will consist of applying a fixed IPv6 40 octets header
containing source and destination address as described above, the
copy of original flow label, the copied and decremented hop limit
count and, depending on the local policy, CoS setting (copy of
original or setting local value). In the case of IPv4 scenario the
20 octets IP header will contain TTL copied and decremented from
original packet, CoS (copy of original or setting of local value) + 8
octets UDP header allowing to improve entropy of flows bundled to
travel within the provided TE path yet to still be able to utilize
any ECMP along the path list.
o Encapsulated packets are natively forwarded via the network (by
and through P nodes) till they arrive at the destination Segment
Endpoint where the destination address gets swapped to the new
destination address from the PATH_LIST kept in the local control
and data plane. The lookup which returns new destination of the
packet is a source-destination based lookup using both
PATH_GID_PFX (with PATH_GID being encoded in the least significant
Raszuk Expires March 29, 2020 [Page 9]
Internet-Draft IP TE+NP September 2019
bits of the source address of the packet) and SID (encoded in
destination address of the packet). That allows to maintain very
good scaling property of the solution without SID state or SID
number explosion. All functions descriptions which are encoded in
the SIDs can be reused across any segment endpoint, if required,
as they have only local significance.
o When packets arrive at the destination PE (last Segment Node) a
similar lookup is performed which returns NULL as next segment
what in turn will result into the decapsulation of the packet and
regular destination based lookup of the destination address
present in the inner IP header. As noted, a local optimization
allows to encode the local lookup type in last 4 bits of any SID
hence allowing to skip the first lookup if such optimization is
enabled by the operator.
o The described lookup table is instantiated and maintained by
either the control plane or by the local configuration of sets of
path lists. For any given segment end node, only local SIDs
(those where most significant prefix bits match locally configured
prefixes) are populated to data plane along with PATH_GIDs they
are attached to. That setup is all what is required to provide
basic IP TE service. More elaboration on other SID values will be
described within the embedded network programming section below.
5. Control plane
The proposed solution is based on classic IP reachability and does
not require any new control plane extension. In its basic form, and
in order to setup a few TE paths across the sample network in
Figure 1, all is required is to apply two path lists on ingress and
egress nodes as well as on three segment endpoints.
However depending on the required TE scale, on the network size, as
well as on the TE path complexity, real production deployments will
likely utilize automation in order to provision such configurations.
Local NMS can be used successfully to provision all participating
segment nodes with proper set of path lists. A separate document
specification describing yang models for the solution will be
provided.
Another alternative to propagate set of path lists can be enabled by
using segment routing extensions for PCEP as described in
[I-D.ietf-pce-segment-routing]. For the basic TE use cases path
lists used are identical to SID lists for SR-MPLS or SRv6
technologies. The logic used by PCE to compute such paths within
given domain can be directly leveraged by this architecture. The
defined SR-ERO sub-object can be directly used to propagate path
Raszuk Expires March 29, 2020 [Page 10]
Internet-Draft IP TE+NP September 2019
lists not also to ingress and egress nodes, but also to all segment
end points participating in given path list transit.
The described above methods offer a manual or automated way to
distribute path lists from central locations using directed TCP
sessions to all participating network elements. However, in order to
even further reduce the complexity and increase rate of path list
propagation across any domain a point to multipoint solution could be
utilized. Also here like in former cases, existing extensions are
available - specifically extension to BGP in order to Advertise
Segment Routing Policies as described in
[I-D.ietf-idr-segment-routing-te-policy]. Detailed encoding examples
will be provided in subsequent versions of this document.
BGP constructs used for SR Policies propagation to ingress nodes can
be used as is in order to propagate analogues path lists to all
participating nodes in the network. A new SAFI has been defined
(codepoint 73) to separate such propagation from any other address
family as well as to uniquely define the NLRI format. For the
purpose of dissemination path lists NLRI 4 octet Policy Color will
carry CLASSIFIER_ID and 4 or 16 octet Endpoint field will carry the
PATH_GID value. If PATH_GID is shorter then 4 or 16 octets the most
significant bits of Endpoint field will be set to zero. Ordered list
of SIDs will be propagated using Segment List Sub-TLVs (Type 3 for
IPv4 and Type 9 for IPv6). Optionally other Sub-TLVs can be also
included with propagation of path lists - for example: Preference
Sub-TLV, Priority Sub-TLV, Name Sub-TLV etc...
As intra-domain BGP usually employs route reflection it is likely
that participating nodes may receive many more path lists then
required to be kept or installed into data plane. There are two
optional solutions to reduce amount of unnecessary control plane
information required to be kept any participating node which when
applied on ingress will result in path lists inbound filtering: use
of route target extended communities or filtering based on
intersection of locally configured IP prefixes with either prefix
part of Endpoint NLRI or prefix part of any SID carried in Segment
List Sub-TLVs. Even if all path lists received would be accepted by
BGP for operational and troubleshooting needs only those which are
locally significant will be installed into data plane.
6. Data plane
There are three IP TE+NP deployment scenarios which may require
different data plane encoding specific to the type of connectivity
available for ingress, egress and TE transit nodes. The following
three categories are covered by this specification:
Raszuk Expires March 29, 2020 [Page 11]
Internet-Draft IP TE+NP September 2019
Cat I - deployment within service provider or enterprise where all
participating nodes are interconnected via links operated by the
same organization using addressing scheme in control of such
organization
Cat II - deployment where participating sites are interconnected
over third party operated networks, where participating in IP TE
nodes could allocate sufficient address block to be used as source
address and still permit to encode entire PATH_GID space of the
size chosen by the operator in the least significant bits of the
addresses of such nodes
Cat III - deployment where participating nodes are interconnected
over third party operated infrastructure where all what has been
granted to such nodes are either host routes or prefixes with not
enough bits left to encode PATH_GID
The below building blocks constitute the required minimum data plane
functionality for this architecture:
Source+Destination Routing [I-D.ietf-rtgwg-dst-src-routing]
Choice of encapsulation:
IPv4 in IPv4+UDP [I-D.xu-intarea-ip-in-udp]
IPv6 or IPv4 in IPv6 [RFC2473]
The selection of normal destination only lookup or source+destination
lookup is triggered by lookup of the destination address. Network
elements which do not participate in the IP TE+NP service will
perform destination only lookup and forward the packets. Network
elements which do participate in the new architecture will perform
destination address check and if that address matches the local
prefix assigned to IP TE+NP service source+destination lookup will
take place, otherwise standard destination only lookup will be
performed.
For deployments falling into Cat III as classified above available
address space does not allow to encode the PATH_GID as part of the
source address. Therefore in such scenarios it is recommended to use
additional GRE encapsulation where PATH_GID would be encoded in the 4
octet key field.
Proposed above GRE header encoding applicable only to Cat III
deployments should in addition to already defined rules also follow
described GRE encoding in the following specifications:
Raszuk Expires March 29, 2020 [Page 12]
Internet-Draft IP TE+NP September 2019
IPv4 in IPv4+UDP+GRE [RFC8086]
IPv4 or IPv6 in IPv6+GRE [RFC7676]
In Cat III deployments when source+destination lookup is performed
PATH_GID from GRE key field should be used instead of packet's source
address. For the case of IPv6 packet encapsulation 12 octets of
zeros should be locally prepended to the key to perform
source+destination lookup.
7. Network Programming
Control Plane Assisted Traffic Engineering is fully compatible with
functions as described in [I-D.ietf-spring-srv6-network-programming]
with one major difference. Instead of always inserting SIDs in a
form of SRH on ingress and into each packet, there are few
alternative ways proposed by this specification. One of them assumes
that information about selected functions is added to the packet by
the penultimate node of a given segment end node hop. SIDs defined
in this document consist of routable prefix part and locally
significant function/instruction part with optional parameters and
lookup type. They can be 32 bit in the case of IPv4 or 128 bit long
in the case of IPv6 with the length of the routable part being a
local choice of the operator.
PATH_GID+SID lookup can return a simple pointer to the next segment
node or can also result in any other local packet processing chain.
While the routable part of the SID has domain-wide significance the
function part has only local meaning to a given node on which it has
been instantiated.
It needs to be observed that some network functions can, for
practical purposes, only be instantiated of the ingress to the domain
and as such can be attached to the packet during initial
encapsulation by use of Segment Routing Header (SRH). The examples
of such functions include L3VPN or EVPN or L2VPN demux labels which
are to be used when packets arrive to the other side of the domain
with or without TE.
To further simplify the processing of packets via the segment end
nodes and relax the requirement for each transit node to inspect SRH
(when added by ingress node) the document will recommend that each
operator in the domain will reserve the last 4 bits of the SID to
explicitly indicate the required lookup type (aka switching vector)
on the outer packet header to occur:
Raszuk Expires March 29, 2020 [Page 13]
Internet-Draft IP TE+NP September 2019
+---------------+---------------------------------+
| Decimal value | Lookup Type |
+---------------+---------------------------------+
| 0 | SRC-DST lookup only |
| 1 | SRH inspection + SRC-DST lookup |
| 2 | Decapsulation + Global lookup |
| 3 | SRH inspection + Decapsulation |
| 4 | reserved |
| .. | .. |
| 15 | reserved |
+---------------+---------------------------------+
Table 1: Recommended allocation of domain wide IPv6 SID_PFX actions
As this specification is only of informational category the proposed
recommendation has non binding character and can be locally replaced
by any different schema as chosen by the operator and made possible
by implementations. For example the 4 bits may be placed in any
other offset after the SID's routable prefix part. The proposed SID
Lookup Types do not replace or interfere in any way with SRH SRv6
Endpoint Behaviors as defined in
[I-D.ietf-spring-srv6-network-programming].
As defined today [RFC8200] mandates to inspect and process all
extension headers in the IPv6 packet when packet's destination
matches any of the locally configured IPv6 address. Therefor if
present SRH will need to be inspected and processed at each segment
end even if it is known by control plane that it does not contain any
instructions to be executed at a given network element ahead of time.
Authors will however still encourage recommended SID structure to be
used for either troubleshooting reasons or for the future when IPv6
specification will relax the EH handling rules to accomodate such new
deployment models.
As an alternative solution to avoid unnecessary processing of
extension header by nodes which are not required to do so
implementation can treat SID with last four bits set to zero as none
local destination address. In such scenario source+destination
lookup will instead of triggering local extension header processing
invoke destination IPv6 NAT function as defined in [RFC6296]. The
NAT rules which will be pre-programmed using information contained in
the PATH_LIST will effectively result in destination address swap.
Such NAT translation is to be of unidirectional character can can
remain fully stateless.
Described solution also directly applies to the case of IPv4 in IPv6
encapsulation.
Raszuk Expires March 29, 2020 [Page 14]
Internet-Draft IP TE+NP September 2019
In the case of IPv4 in IPv4+UDP encapsulation the basic behaviour of
embedding functions in SIDs does not change. However as to the
moment of this writing the proposed IPv4 header extensions
[I-D.herbert-ipv4-eh] and [I-D.herbert-ipv4-udpencap-eh] may only
allow limited number of extension headers to be used (Hop-by-Hop
Options and Destination Options). As such the recommended allocation
table in the case of IPv4 requires slight adjustment:
+---------------+---------------------------------+
| Decimal value | Lookup Type |
+---------------+---------------------------------+
| 0 | SRC-DST lookup only |
| 1 | DOH inspection + SRC-DST lookup |
| 2 | Decapsulation + Global lookup |
| 3 | DOH inspection + Decapsulation |
| 4 | reserved |
| .. | .. |
| 15 | reserved |
+---------------+---------------------------------+
Table 2: Recommended allocation of domain wide IPv4 SID_PFX actions
The specific syntax of Destination Option Header encoding when used
with IPv4 encapsulation will be defined in subsequent versions of
this document.
Existing services (ex: MPLS-VPNs [RFC4364]) are fully compatible as-
is without any modifications to be transported over described IP TE
architecture. Existing MPLS label can be used as service demux with
full replacement of MPLS-Transport to IP-TE transport. In such
scenario there is no longer need to rename service demux value into
some new nomenclature to artificially force it to fit into SID space.
Substitute of MPLS transport with new IP TE transport is essentially
treated as basic IP-in-IP encapsulation and is seamless to the upper
layer applications. That however in no way can prevent invention of
new native services to only use new network programming paradigm.
8. Active Path Probing
One of the critical network metrics for a lot of applications running
on the network is not only ability to reach the destination in a
relatively congestion free fashion, but also the quality of the path
which is traversed towards a destination. The latter is,
unfortunately, very seldom used as selection criteria in number of TE
implementations. Here authors recommend that, from day one, the
operator has an option in order to define the minimum path quality
metrics before it is considered for actual data plane use as both
Raszuk Expires March 29, 2020 [Page 15]
Internet-Draft IP TE+NP September 2019
relative or absolute set of values. Comparison with non TE path or
other TE paths end to end metrics should also be available.
Today's network technologies focus on local protection as reaction to
adjacent link or node failures. At the same time, there is a
significant concern that they lack detection of any malfunctions of
network elements' internal data plane itself which, as proven in
number of production deployments, does occur.
Moreover, it also needs to be observed that most if not all of
commonly used routing protocols focus on assuring loop free
destination reachability via shortest or best path measured with
static metrics without any consideration given to actual quality of
end to end path towards given destination.
Traffic engineering allows to enable real time SLA evaluation of
various TE paths. Results of such measurements can be used to
automatically map traffic to such TE transport. Architecture
described by this specification integrates such functionality
provided an operator chooses to enable it.
It needs to be noted that packets used for diagnostics must traverse
the exact same data plane and should be encapsulated in the identical
header as the user packets. Such measurements not only detect path
parameters but also end to end path availability.
While (N times path RTT - N times local detection interval) slower
from local protection for vast majority of applications such end to
end path liveness detection rate is both sufficient for applications
and much simpler to implement and operate. It is also more
attractive due to increased spectrum of types of failures which can
be detected. Removed complexity required to be employed (example:
node protection repair of adjacent segment nodes) is also an
important consideration.
The choice of path probing protocol is left as the local operator's
decision. However, it needs to be observed that such protocol suite
should allow fast liveness detection as well as end to end path
quality measurements reported to path headend (typically a network
ingress node) as RTT, Jitter, Delay, MOS parameters as well as max
MTU and sweep MTU path validation.
It is also completely valid to use more than one protocol - each in
different frequency setting. As an example, one could use BFD
multihop [RFC5883] with hardware offload to detect end to end path
liveness while in the same time apply OWAMP [RFC4656] to collect more
unidirectional path quality metrics. Recommendation for a single
Raszuk Expires March 29, 2020 [Page 16]
Internet-Draft IP TE+NP September 2019
integrated path liveness and quality reporting protocol will also be
described in a separate IETF specification.
8.1. TI-LFA Local Protection
As stated in the TI-LFA specification for networks supporting segment
routing [I-D.ietf-rtgwg-segment-routing-ti-lfa], protection of SR
policy midpoints involves adjustments to segment list carried in the
packets as well as proper selection of repair path in order to assure
that protected packets can successfully reach the next SR policy
segment node.
Based on the control plane distribution of complete PATH_LIST,
similar protection is possible in the described architecture.
Without any additional requirements to adjust any other fields in the
packet header only destination address can be swapped. Current
destination can be replaced by subsequent node's destination address
on the PATH_LIST upon detection of neighboring node failure. That
operation however, requires to maintain per path state at PLRs what
while certainly possible may not be operator's preference.
Enabling local protection in segment engineered IP networks is
clearly possible, however it needs additional processing and control
plane information to be distributed and present on all nodes in the
domain. Protection PATH_LISTs can be either computed centrally or by
any node in the domain (including PLRs). Authors recommend this to
remain a local operator decision and at the same time encourage to
use end to end path protection scheme as first preference.
9. Solution advantages
The following key advantages can be used to characterize the
described architecture:
o Native TE support for IPv4 and IPv6
o Very efficient use of available address space - no requirement for
any new address allocations
o IGP impact - single prefix injection from ingress nodes of length
chosen by operator
o Ability to aggregate injected prefixes at area or domain boundary
with no impact to functionality
o No extensions to ISIS or OSPF routing protocols required
Raszuk Expires March 29, 2020 [Page 17]
Internet-Draft IP TE+NP September 2019
o Reuse of commonly available components (SRC-DST routing and IPinIP
encapsulation)
o Integrated end to end path validation for reachability and quality
o For basic TE and PATH_LIST SID integrated network programming
functions fixed overhead of 28 octets for IPv4 and 40 octets for
IPv6.
o Full compatibility with SRH from SRv6 Network Programming concept
o No per user data flow state in any network element of the network
except ingress (mapping only)
o No packet header size growth with the growing number of TE segment
endpoints policies
o Support in all available hardware - no need for any new operations
on the packet headers
o TI-LFA support when end to end path protection will not be
sufficient
o Full native support of network services: L2VPNs, L3VPNs, EVPNs etc
with single SID in SRH or native service level encapsulation
o Support of ingress, egress or transit nodes with available only
single host address available on each such system
10. OAM
As result of use of IP encapsulation both traceroute as well as ping
are natively supported within a given domain boundaries. ICMP or UDP
OAM probes will be encapsulated in the exact same IPv4 or IPv6 header
as user data packets therefore all replies will be sent to the domain
ingress node.
No modifications to additional extension headers or even their
presence is required for correct OAM operations.
If an OAM packet is originated externally to the domain, the ingress
node will need to act as OAM proxy in relaying the responses to its
original sources.
Raszuk Expires March 29, 2020 [Page 18]
Internet-Draft IP TE+NP September 2019
11. Deployment considerations
The solution is defined to be fully customizable by the operator.
The path engineering as well as choice of numbering will likely
differ domain to domain.
As all packets subject to this specification carry in their source
address immutable PATH_GID. Together with locally assigned SIDs no
further extensions are necessary to identify specific path flows at
any point in the domain. The same tuple PATH_GIDs + SIDs can also be
used to identify any path statistics (netflow records) at any point
in the domain.
12. Security considerations
No new security issues are introduced by this specification.
13. IANA Considerations
No IANA allocations are required by this specification.
14. Acknowledgements
Authors would like to thank Tony Li, Stefano Previdi, Dirk Steinberg
and Francois Clad for their valuable review and comments.
15. References
15.1. Normative References
[RFC1918] Rekhter, Y., Moskowitz, B., Karrenberg, D., de Groot, G.,
and E. Lear, "Address Allocation for Private Internets",
BCP 5, RFC 1918, DOI 10.17487/RFC1918, February 1996,
<https://www.rfc-editor.org/info/rfc1918>.
[RFC2003] Perkins, C., "IP Encapsulation within IP", RFC 2003,
DOI 10.17487/RFC2003, October 1996,
<https://www.rfc-editor.org/info/rfc2003>.
[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>.
[RFC2473] Conta, A. and S. Deering, "Generic Packet Tunneling in
IPv6 Specification", RFC 2473, DOI 10.17487/RFC2473,
December 1998, <https://www.rfc-editor.org/info/rfc2473>.
Raszuk Expires March 29, 2020 [Page 19]
Internet-Draft IP TE+NP September 2019
[RFC2784] Farinacci, D., Li, T., Hanks, S., Meyer, D., and P.
Traina, "Generic Routing Encapsulation (GRE)", RFC 2784,
DOI 10.17487/RFC2784, March 2000,
<https://www.rfc-editor.org/info/rfc2784>.
[RFC4364] Rosen, E. and Y. Rekhter, "BGP/MPLS IP Virtual Private
Networks (VPNs)", RFC 4364, DOI 10.17487/RFC4364, February
2006, <https://www.rfc-editor.org/info/rfc4364>.
[RFC6296] Wasserman, M. and F. Baker, "IPv6-to-IPv6 Network Prefix
Translation", RFC 6296, DOI 10.17487/RFC6296, June 2011,
<https://www.rfc-editor.org/info/rfc6296>.
[RFC7676] Pignataro, C., Bonica, R., and S. Krishnan, "IPv6 Support
for Generic Routing Encapsulation (GRE)", RFC 7676,
DOI 10.17487/RFC7676, October 2015,
<https://www.rfc-editor.org/info/rfc7676>.
[RFC8086] Yong, L., Ed., Crabbe, E., Xu, X., and T. Herbert, "GRE-
in-UDP Encapsulation", RFC 8086, DOI 10.17487/RFC8086,
March 2017, <https://www.rfc-editor.org/info/rfc8086>.
[RFC8126] Cotton, M., Leiba, B., and T. Narten, "Guidelines for
Writing an IANA Considerations Section in RFCs", BCP 26,
RFC 8126, DOI 10.17487/RFC8126, June 2017,
<https://www.rfc-editor.org/info/rfc8126>.
[RFC8174] Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC
2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174,
May 2017, <https://www.rfc-editor.org/info/rfc8174>.
[RFC8200] Deering, S. and R. Hinden, "Internet Protocol, Version 6
(IPv6) Specification", STD 86, RFC 8200,
DOI 10.17487/RFC8200, July 2017,
<https://www.rfc-editor.org/info/rfc8200>.
15.2. Informative References
[I-D.herbert-ipv4-eh]
Herbert, T., "IPv4 Extension Headers and Flow Label",
draft-herbert-ipv4-eh-01 (work in progress), May 2019.
[I-D.herbert-ipv4-udpencap-eh]
Herbert, T., "IPv4 Extension Headers and UDP Encapsulated
Extension Headers", draft-herbert-ipv4-udpencap-eh-01
(work in progress), March 2019.
Raszuk Expires March 29, 2020 [Page 20]
Internet-Draft IP TE+NP September 2019
[]
Filsfils, C., Dukes, D., Previdi, S., Leddy, J.,
Matsushima, S., and d. daniel.voyer@bell.ca, "IPv6 Segment
Routing Header (SRH)", draft-ietf-6man-segment-routing-
header-23 (work in progress), September 2019.
[I-D.ietf-idr-segment-routing-te-policy]
Previdi, S., Filsfils, C., Mattes, P., Rosen, E., Jain,
D., and S. Lin, "Advertising Segment Routing Policies in
BGP", draft-ietf-idr-segment-routing-te-policy-07 (work in
progress), July 2019.
[I-D.ietf-pce-segment-routing]
Sivabalan, S., Filsfils, C., Tantsura, J., Henderickx, W.,
and J. Hardwick, "PCEP Extensions for Segment Routing",
draft-ietf-pce-segment-routing-16 (work in progress),
March 2019.
[I-D.ietf-rtgwg-dst-src-routing]
Lamparter, D. and A. Smirnov, "Destination/Source
Routing", draft-ietf-rtgwg-dst-src-routing-07 (work in
progress), March 2019.
[I-D.ietf-rtgwg-segment-routing-ti-lfa]
Litkowski, S., Bashandy, A., Filsfils, C., Decraene, B.,
Francois, P., daniel.voyer@bell.ca, d., Clad, F., and P.
Camarillo, "Topology Independent Fast Reroute using
Segment Routing", draft-ietf-rtgwg-segment-routing-ti-
lfa-01 (work in progress), March 2019.
[I-D.ietf-spring-srv6-network-programming]
Filsfils, C., Camarillo, P., Leddy, J.,
daniel.voyer@bell.ca, d., Matsushima, S., and Z. Li, "SRv6
Network Programming", draft-ietf-spring-srv6-network-
programming-03 (work in progress), September 2019.
[I-D.patel-raszuk-bgp-vector-routing]
Raszuk, R., Patel, K., Pithawala, B., Sajassi, A.,
Osborne, E., Jalil, L., and J. Uttaro, "BGP vector
routing.", draft-patel-raszuk-bgp-vector-routing-07 (work
in progress), May 2016.
[I-D.xu-intarea-ip-in-udp]
Xu, X., Assarpour, H., Ma, S., daniel.bernier@bell.ca, d.,
Dukes, D., Lee, Y., and F. Yongbing, "Encapsulating IP in
UDP", draft-xu-intarea-ip-in-udp-07 (work in progress),
May 2018.
Raszuk Expires March 29, 2020 [Page 21]
Internet-Draft IP TE+NP September 2019
[RFC4655] Farrel, A., Vasseur, J., and J. Ash, "A Path Computation
Element (PCE)-Based Architecture", RFC 4655,
DOI 10.17487/RFC4655, August 2006,
<https://www.rfc-editor.org/info/rfc4655>.
[RFC4656] Shalunov, S., Teitelbaum, B., Karp, A., Boote, J., and M.
Zekauskas, "A One-way Active Measurement Protocol
(OWAMP)", RFC 4656, DOI 10.17487/RFC4656, September 2006,
<https://www.rfc-editor.org/info/rfc4656>.
[RFC5883] Katz, D. and D. Ward, "Bidirectional Forwarding Detection
(BFD) for Multihop Paths", RFC 5883, DOI 10.17487/RFC5883,
June 2010, <https://www.rfc-editor.org/info/rfc5883>.
[RFC8402] Filsfils, C., Ed., Previdi, S., Ed., Ginsberg, L.,
Decraene, B., Litkowski, S., and R. Shakir, "Segment
Routing Architecture", RFC 8402, DOI 10.17487/RFC8402,
July 2018, <https://www.rfc-editor.org/info/rfc8402>.
Author's Address
Robert Raszuk (editor)
Bloomberg LP
731 Lexington Ave
New York City, NY 10022
USA
Email: robert@raszuk.net
Raszuk Expires March 29, 2020 [Page 22]