| Internet-Draft | APN Scope and Gap Analysis | March 2022 |
| Peng, et al. | Expires 8 September 2022 | [Page] |
The APN work in IETF is focused on developing a framework and set of mechanisms to derive, convey and use an attribute allowing the implementation of fine-grain user group-level and application group-level requirements in the network layer. APN aims to apply various policies in different nodes along a network path onto a traffic flow altogether, for example, at the headend to steer into corresponding path, at the midpoint to collect corresponding performance measurement data, and at the service function to execute particular policies. Currently there is still no way to efficiently realize this composite network service provisioning along the path. This document further clarifies the scope of the APN work and describes the solution gap analysis.¶
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 8 September 2022.¶
Copyright (c) 2022 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 Revised BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Revised BSD License.¶
Application-aware Networking (APN) is introduced in [I-D.li-apn-framework] and [I-D.li-apn-problem-statement-usecases]. APN conveys an attribute along with data packets into network and makes the network aware about data flow requirements at different granularity levels.¶
Such an attribute is acquired, constructed in a structured value, and then encapsulated in the packet. Such structured value is treated as an opaque object in the network to which the network operator applies policies in various nodes/service functions along the path and provides corresponding services.¶
This structured attribute can be encapsulated in various data planes adopted within a Network Operator controlled limited domain, e.g. MPLS, VXLAN, SR/SRv6 and other tunnel technologies, which waits to be further specified.¶
With APN, it becomes possible to apply various policies in different nodes along a network path onto a traffic flow altogether in a more efficient way, e.g., at the headend to steer into corresponding path, at the midpoint to collect corresponding performance measurement data, and at the service function to execute particular policies. Currently there is still no way to realize this composite network service provisioning along the path very efficiently. It may be possible to stack those various policies in a list of TLVs at the headend. However, this approach would introduce great complexities and impose big challenges on the hardware processing and forwarding.¶
The example use-case presented in this draft further expands on the rationale for such an attribute and how it can be derived and used in that specific context.¶
This document further clarifies the scope of the APN work and describes the solution gap analysis.¶
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14 RFC 2119 [RFC2119] RFC 8174 [RFC8174] when, and only when, they appear in all capitals, as shown here.¶
APN: Application-aware Networking¶
CPE: Customer Premises Equipment¶
DPI: Deep Packet Inspection¶
OS: Operating System¶
The APN framework is introduced in [I-D.li-apn-framework], as shown in the Figure 1.¶
+-----+ +-----+
|App x|-\ /-|App x|
+-----+ | +-----+ +-----------------------+ +-----+ | +-----+
\-|APN- | | Application-aware | |APN- |-/
| |---| Network |---| |
/-|Edge | | Service Provisioning | |Edge |-\
+-----+ | +-----+ +-----------------------+ +-----+ | +-----+
|App y|-/ | | \-|App y|
+-----+ |<--- Network Operator Controlled --->| +-----+
Limited Domain
Figure 1. APN Framework and Scope
¶
APN is only applied to an edge-to-edge tunnel encapsulation within a limited trusted domain. It means that the source and destination addresses of the packet are the endpoints of the tunnel (i.e. the domain edges), and nothing about the payload source and destination can be deduced, which substantially reduces the privacy concerns. Typically, an APN domain is defined as a Network Operator controlled limited domain (see Figure 1), in which MPLS, VXLAN, SR/SRv6 and other tunnel technologies are adopted to provide network services.¶
With APN, the attribute is acquired based on the existing information in the packet header (i.e. source and destination addresses, incoming L2 (or) MPLS encapsulation, incoming physical/virtual port information, the other fields of the 5-tuple if they are not encrypted) at the edge devices of the APN domain, added to the data packets along with the tunnel encapsulation, and delivered to the network, wherein, according to this attribute, corresponding network services are provisioned. When the packets leave the APN domain, the attribute is removed together with the tunnel encapsulation header.¶
To be more specific and more concrete, here we use SD-WAN as an example use case to further expand on the rationale for such attribute and how it can be derived and used in that specific context.¶
In the case of SD-WAN, an enterprise obtains WAN services from an SD-WAN provider so that its employees have access to the applications in the Cloud, and then the SD-WAN provider may buy WAN lines from a Network Operator. The enterprise may know what applications will use the SD-WAN services, but it will only provide the 5 tuples (i.e. source IP address, source port, destination IP address, destination port, transport protocol) of those applications to the SD-WAN provider. So, the SD-WAN provider does not know what applications it is serving, and will only provide 5 tuples to the Network Operator and the service performance requirements for steering their customer's traffic. In this way, the Network Operator does not know anything else about the traffic except the 5 tuples and requirements. Nowadays, SD-WAN is usually using 5-tuple to steer the traffic into corresponding WAN lines across the Network Operator's network [SD-WAN].¶
However, there are two main issues in the current SD-WAN deployments.¶
1) It is complicated to resolve the 5 tuples. Even worse, as the traffic is encrypted, it becomes impossible to obtain any transport layer information. Moreover, in the IPv6 data plane, with the extension headers being added before the upper layer, in some implementations it becomes very difficult and even impossible to obtain transport layer information because that information is located deep in the packet. So, there is no 5 tuples anymore, and maybe only 2 tuples are available.¶
2) Currently there is still no way to apply various policies in different nodes along the network path onto a traffic flow altogether, that is, at the headend to steer into corresponding path, at the midpoint to collect corresponding performance measurement data, and at the service function to execute particular policies. It may be possible to stack those various policies in a list of TLVs at the headend. However, this approach would introduce great complexities and impose big challenges on the hardware processing and forwarding.¶
With APN, at the edge node, i.e. CPE, of the SD-WAN (see Figure 2), the 5-tuple, plus information related to user or application group-level requirements is constructed into a structured value, called APN attribute. This attribute is only meaningful for the network operators to apply various policies in different nodes/service functions, which can be enforced from the Controllers.¶
+-----------------+
+---------|SD-WAN Controller|---------+
| +--------|--------+ |
| | |
| +-------|-------+ |
| |SDN Controller| |
| +-------|-------+ |
+-----+ | | | +-----+
|App x|-\ | | | /-|App x|
+-----+ | +--|--+ +-----------|-----------+ +--|--+ | +-----+
\-| | | Application-aware | | |-/
|CPE 1|---| Network |---|CPE 2|
/-| | | Service Provisioning | | |-\
+-----+ | +-----+ +-----------------------+ +-----+ | +-----+
|App y|-/ | | \-|App y|
+-----+ |<--- Network Operator Controlled --->| +-----+
Limited Domain
Figure 2. SD-WAN using the APN Framework
¶
With such an attribute in the network, we can easily solve the two issues above-mentioned. For example, when the packet is sent from the CPE1 and the attribute is added along with the tunnel encapsulation, then it is not necessary to resolve the 5-tuple and perform the deep inspection in every node along the path. This attribute is encapsulated in the network layer and can be easily read by the routers and service functions. If the tunnel is based on the IPv6 data plane, for example, such an attribute can be encapsulated in an option of IPv6 hop-by-hop options header.¶
Since this attribute is taken as an object to the network, the network operators will simply place the policies in the nodes/service functions where this indicated traffic will go through, and the corresponding node/service function will just apply policies for this object. This can be easily done by utilizing this attribute, which is not possible with any current existing mechanism.¶
Such attribute will also bring other benefits, for example,¶
Furthermore, with such attribute, more new services could be enabled, for example,¶
There are already some solutions specified in IETF, which use identifier to perform traffic steering and service provisioning. However, the existing solutions are specific to a particular scenario or data plane. None of them is the same as APN and able to achieve the same effects.¶
[RFC6437] specifies the IPv6 flow label which enables the IPv6 flow classification. However, the IPv6 flow label is mainly used for Equal Cost Multipath Routing (ECMP) and Link Aggregation [RFC6438].¶
Similarly, [RFC6391] describes a method of adding an additional Label Stack Entry (LSE) at the bottom of the stack in order to facilitate the load balancing of the flows within a pseudowire (PW) over the available ECMPs. A similar design for general MPLS use has also been proposed in [RFC6790] using the concept of Entropy Label.¶
Subscriber Identifier and Performance Policy Identifier are specified in [RFC8979]. These identifiers are carried only in the Network Service Header (NSH) [RFC8300] Context Header, as shown in Figure 3, while the APN attribute can be carried in various data plane encapsulations.¶
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Metadata Class | Type |U| Length |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| |
~ Subscriber Identifier ~
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Metadata Class | Type |U| Length |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| |
~ Performance Policy Identifier ~
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Figure 3. Subscriber Identifier and Performance Policy Identifier
¶
In this draft [RFC8979], the Subscriber Identifier carries an opaque local identifier that is assigned to a subscriber by a network operator, and the Performance Policy Identifier represents an opaque value pointing to specific performance policy to be enforced. In this way, in order to apply various policies in different nodes along the network path onto a traffic flow altogether, e.g., at the headend to steer into corresponding path, at the midpoint to collect corresponding performance measurement data, and at the service function to execute particular policies, those various policies would have to be stacked in a list of TLVs at the headend, introducing great complexities and big challenges on the hardware processing and forwarding.¶
The APN attribute is treated as an opaque object in the network, to which the network operator applies policies in various nodes/service functions along the path and provide corresponding services.¶
A 32-bit Flow ID is specified in [I-D.ietf-ippm-ioam-direct-export], which is used to correlate the exported data of the same flow from multiple nodes and from multiple packets, while the APN attribute can serve more various purposes.¶
The Binding SID (BSID) [RFC8402] is bound to an SR Policy, instantiation of which may involve a list of SIDs. Any packets received with an active segment equal to BSID are steered onto the bound SR Policy. A BSID may be either a local or a global SID. While the APN attribute is not bound to SR only, and it can be carried in various data plane encapsulations.¶
The flow specification (FlowSpec) [RFC5575] is actually an n-tuple consisting of several matching criteria that can be applied to IP traffic, which include elements such as source and destination address prefixes, IP protocol, and transport protocol port numbers. In BGP VPN/MPLS networks, BGP FlowSpec can be extended to identify and change (push/swap/pop) the label(s) for traffic that matches a particular FlowSpec rule in [I-D.ietf-idr-flowspec-mpls-match] and [I-D.ietf-idr-bgp-flowspec-label]. In [I-D.liang-idr-bgp-flowspec-route], BGP is used to distribute the FlowSpec rule bound with label(s). While the APN attribute is not bound to MPLS only, and it can be carried in various data plane encapsulations.¶
The capabilities of the VXLAN-GPE protocol can be extended by defining next protocol "shim" headers that are used to implement new data plane functions. For example, Group Policy ID is carried in the Group-Based Policy (GBP) Shim header [I-D.lemon-vxlan-lisp-gpe-gbp]. GENEVE has similar ability as VXLAN-GPE to carry metadata.¶
Identification and Specification of DetNet Flows is specified in [RFC9016]. DetNet MPLS flows can be identified and specified by the SLabel and the FLabelStack. The IP 6-tuple is used for DetNet IP flow identification, which consists of SourceIpAddress, DestinationIpAddress, Dscp, Protocol, SourcePort, and DestinationPort. IPv6FlowLabel and IPSecSpi are additional attributes that can be used for DetNet flow identification in addition to the 6-tuple. Therefore, the Detnet IP Flow ID is logical and there is no such Flow ID carried for Detnet, but only the 6-tuple is directly used to identify the Detnet flows.¶
Only one exceptional case, in [I-D.ietf-spring-sr-redundancy-protection], the 32-bit flow identification (FID) identifies one specific Detnet flow of redundancy protection. This FID is usually allocated from centralized controller to the SR ingress node or redundancy node in SR network.¶
In [I-D.dong-6man-enhanced-vpn-vtn-id], VTN Resource ID is a 4-octet identifier which uniquely identifies the set of network resources allocated to a VTN. For network slicing, the ID is used to indicate the network resources to be allocated to the network slices and it is not bound to any traffic flow.¶
APN is for traffic steering, while network slicing is about resource partition [I-D.ietf-teas-rfc3272bis].¶
In [RFC8300], Service Path Identifier (SPI) uniquely identifies a Service Function Path (SFP). Participating nodes MUST use this identifier for SFP selection. The initial Classifier MUST set the appropriate SPI for a given classification result. For SFC, the ID is used to indicate a SF path and it is not bound to any traffic flow.¶
The comparison of the identifiers for the typical network services (incl. iOAM, Detnet, Network Slicing (NS), and Service Function Chaining (SFC)) is shown in the following Table from different aspects (incl. ID, Identification Object, Source (for generating the ID), Configuration (Conf.) node, and Size).¶
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| | ID | Identification Object | Source |Conf. node| Size |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| APN | APN ID | The flow that needs | 5-tuple|Controller|32bits|
| | | fine-granular services| Layer 2| |128b |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| iOAM | Flow ID | The flow that needs | - |Controller|32bits|
| | | performance monitoring| | Ingress | |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|Detnet| Flow ID | The flow that needs | - |Controller| - |
| | (6-tuple) | Detnet services | | | |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|Detnet| Flow ID | The redundant | - | Detnet |32bits|
| | | protection flow | |Controller| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| NS |Resource ID| The network resources | |Controller|32bits|
| | | that are allocated to | - | | |
| | | network slices | | | |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| SFC | SPI | The SF Path | - |Controller|24bits|
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| SFC |Performance| The performance policy| - |Controller| - |
| | Policy ID | | | | |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Table 1. Comparison of the Identifiers
¶
As driven by ever-emerging new 5G services, fine-granularity service provisioning becomes urgent. The existing solutions are either specific to a particular scenario or data plane. While APN aims to define a generalized attribute used for fine-granularity service provisioning, and can be carried in various data plane encapsulations.¶
There are no IANA considerations in this document.¶
The authors would like to acknowledge Martin Vigoureux, Alvaro Retana, Barry Leiba, Stefano Previdi, Adrian Farrel, and Daniel King for their valuable review and comments.¶