| Internet-Draft | EIP Architecture | June 2024 |
| Salsano, et al. | Expires 24 December 2024 | [Page] |
Extensible In-band Processing (EIP) extends the functionality of the IPv6 protocol considering the needs of future Internet services / 6G networks. This document discusses the architecture and framework of EIP. Two separate documents respectively analyze a number of use cases for EIP and provide the protocol specifications of EIP.¶
This note is to be removed before publishing as an RFC.¶
The latest revision of this draft can be found at https://eip-home.github.io/eip-arch/draft-eip-arch.html. Status information for this document may be found at https://datatracker.ietf.org/doc/draft-eip-arch/.¶
Discussion of this document takes place on the EIP SIG mailing list (mailto:eip@cnit.it), which is archived at http://postino.cnit.it/cgi-bin/mailman/private/eip/.¶
Source for this draft and an issue tracker can be found at https://github.com/eip-home/eip-arch.¶
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 24 December 2024.¶
Copyright (c) 2024 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.¶
Networking architectures need to evolve to support the needs of future Internet services and 6G networks. The networking research and standardization communities have considered different approaches for this evolution, that can be broadly classified in 3 different categories:¶
Clean slate and "revolutionary" solutions. Throw away the legacy IP networking layer.¶
Solutions above the layer 3. Do not touch the legacy networking layer (IP).¶
Evolutionary solutions. Improve the IP layer (and try to preserve backward compatibility).¶
The proposed EIP (Extensible In-band Processing) solution belongs to the third category, it extends the current IPv6 architecture without requiring a clean-slate revolution.¶
The use cases for EIP are discussed in [id-eip-use-cases]. The specification of the EIP header format is provided in [id-eip-headers].¶
An ongoing trend is extending the functionality of the IPv6 networking layer, going beyond the plain packet forwarding. An example of this trend is the rise of the SRv6 "network programming" model. With the SRv6 network programming model, the routers can implement "complex" functionalities and they can be controlled by a "network program" that is embedded in IPv6 packet headers. Another example is the INT (IN band Telemetry) solution for monitoring. These (and other) examples are further discussed in Section 4.¶
The EIP solution is aligned with this trend, which will ensure a future proof evolution of networking architectures. EIP supports a feature-rich and extensible IPv6 networking layer, in which complex dataplane functions can be executed by end-hosts, routers, virtual functions, servers in datacenters so that services can be implemented in the smartest and more efficient way.¶
The EIP solution foresees the introduction of an EIP header in the IPv6 packet header. The proposed EIP header is extensible and it is meant to support a number of different use cases. In general, both end-hosts and transit routers can read and write the content of this header. Depending of the specific use-case, only specific nodes will be capable and interested in reading or writing the EIP header. The use of the EIP header can be confined to a single domain or to a set of cooperating domains, so there is no need of a global, Internet-wide support of the new header for its introduction. Moreover, there can be usage scenarios in which legacy nodes can simply ignore the EIP header and provide transit to packets containing the EIP header.¶
An important usage scenario considers the transport of user packets over a provider network. In this scenario, we consider the network portion from the provider ingress edge node to the provider egress edge node. The ingress edge node can encapsulate the user packet coming from an access network into an outer packet. The outer packet travels in the provider network until the egress edge node, which will decapsulate the inner packet and deliver it to the destination access network or to another transit network, depending on the specific topology and service. Assuming that the IPv6/SRv6 dataplane is used in the provider network, the ingress edge node will be the source of an outer IPv6 packet in which it is possible to add the EIP header. The outer IPv6 packet (containing the EIP header) will be processed inside the "limited domain" (see [RFC8799]) of the provider network, so that the operator can make sure that all the transit routers either are EIP aware or at least they can forward packets containing the EIP header. In this usage scenario, the EIP framework operates "edge-to-edge" and the end-user packets are "tunneled" over the EIP domain.¶
The architectural framework for EIP is depicted in Figure 1. We refer to nodes that are not EIP capable as legacy nodes. An EIP domain is made up by EIP aware routers (EIP R) and can also include legacy routers (LEG R). At the border of the EIP domain, EIP edge nodes (EIP ER) are used to interact with legacy End Hosts / Servers (LEG H) and with other domains. It is also possible that an End Host / Server is EIP aware (EIP H), in this case the EIP framework could operate "edge-to-end" or "end-to-end".¶
LEG domain
+------------+
+---+ +---+ +---+ +---+
|EIP|_ _|EIP|______|EIP| ___|LEG|
| H | \__+---+__/ | R | | R |__ +---+__/ | R | ...
+---+ |EIP| +---+ +---+ \__|EIP| +---+
__|ER |__ | | __|ER |__
+---+_/ +---+ \_+---+ +---+__/ +---+ \___+---+
|LEG| |LEG|______|LEG| |EIP|
| H | | R | | R | |ER | ...
+---+ +---+ +---+ +---+
+-----------------------------+ +------------+
EIP domain EIP domain
As shown in Figure 1, an EIP domain can communicate with other domains, which can be legacy domains or EIP capable domains.¶
The EIP header will carry different EIP Information Elements that are defined to support the different use cases. There are reasons why it is beneficial to define a common EIP header that supports multiple use cases.¶
The number of available Option Types in HBH header is limited, likewise the number of available TLVs in the Segment Routing Header (SRH) is limited. Defining multiple Option Types or SRH TLVs for multiple use case is not scalable and puts pressure on the allocation of such codepoints. This aspect is further discussed in Section 4.¶
The definition and standardization of specific EIP Information Elements for the different use cases will be simplified, compared to the need of requiring the definition of a new Option Type or SRH TLVs.¶
Different use cases may share a subset of common EIP Information Elements.¶
Efficient mechanism for the processing of the EIP header (both in software and in hardware) can be defined when the different EIP Information Elements are carried inside the same EIP header.¶
In the last few years, we have witnessed important innovations in IPv6 networking, centered around the emergence of Segment Routing for IPv6 (SRv6) [RFC8754] and of the SRv6 "Network Programming model" [RFC8986]. With SRv6 it is possible to insert a Network program, i.e. a sequence of instructions (called segments), in a header of the IPv6 protocol, called Segment Routing Header (SRH).¶
Another recent activity that proposed to extend the networking layer to support more complex functions, concerns the network monitoring. The concept of INT "In-band Network Telemetry" has been proposed since 2015 [onf-int] in the context of the definition of use cases for P4 based data plane programmability. The latest version of INT specifications dates November 2020 [int-spec]. [int-spec] specifies the format of headers that carry monitoring instructions and monitoring information along with data plane packets. The specific location for INT Headers is intentionally not specified: an INT Header can be inserted as an option or payload of any encapsulation type. The In-band Telemetry concept has been adopted by the IPPM IETF Working Group, renaming it "In-situ Operations, Administration, and Maintenance" (IOAM). [RFC9197] discusses the data fields and associated data types for IOAM. The in-situ OAM data fields can be encapsulated in a variety of protocols, including IPv6. The specification details for carrying IOAM data inside IPv6 headers are provided in [RFC9486]. In particular, IOAM data fields can be encapsulated in IPv6 using either Hop-by-Hop Options header or Destination options header.¶
Another example of extensions to IPv6 for network monitoring is specified in [RFC8250], which defines an IPv6 Destination Options header called Performance and Diagnostic Metrics (PDM). The PDM option header provides sequence numbers and timing information as a basis for measurements.¶
The "Alternate Marking Method" is a recently proposed performance measurement approach described in [RFC8321]. [RFC9343] defines a new Hop-by-Hop Option to support this approach.¶
"Path Tracing" [I-D.draft-filsfils-ippm-path-tracing] proposes an efficient solution for recording the route taken by a packet (including timestamps and load information taken at each hop along the route). This solution needs a new Hop-by-Hop Option to be defined.¶
[RFC8558] analyses the evolution of transport protocols. It recommends that explicit signals should be used when the endpoints desire that network elements along the path become aware of events related to trasport protocol. Among the solutions, [RFC8558] considers the use of explicit signals at the network layer, and in particular it mentions that IPv6 hop-by-hop headers might suit this purpose.¶
[RFC9268] specifies a new IPv6 Hop-by-Hop option that is used to record the minimum Path MTU between a source and a destination.¶
The Internet Draft [I-D.draft-ietf-6man-enhanced-vpn-vtn-id] proposes a new Hop-by-Hop option of IPv6 extension header to carry the Network Resource Partition (NRP) information, which could be used to identify the NRP-specific processing to be performed on the packets by each network node along a network path in the NRP.¶
The Internet-Draft [I-D.draft-guan-6man-ipv6-id-authentication] proposes an IPv6 based address label terminal identity authentication mechanism, which uses a new Hop-by-Hop option.¶
The Internet-Draft [I-D.draft-herbert-fast] describes the Firewalls and Service Tickets (FAST) protocol. This is a generic and extensible protocol for hosts to signal network nodes to request services or to gain admission into a network. Tickets are sent in IPv6 Hop-by-Hop options.¶
We have listed several proposals or already standardized solutions that use the IPv6 Hop-by-Hop Options. These Options are represented with a 8 bits code. The first two bits represent the action to be taken if the Options is unknown to a node that receives it, the third bit is used to specify if the content of the Options can be changed in flight. In particular the Option Types that start with 001 should be ignored if unknown and can be changed in flight, which is the most common combination. The current IANA allocation for Option Types starting with 001 is (see https://www.iana.org/assignments/ipv6-parameters/ipv6-parameters.xhtml)¶
32 possible Option Types starting with 001 4 allocated by RFCs - temporary allocated by Internet Drafts 1 allocated for RFC3692-style Experiment 27 not allocated¶
We observe that there is a potential scarcity of the code points, as there are many scenarios that could require the definition of a new Hop-by-hop option. We also observe that having only 1 code point allocated for experiments is a very restrictive limitation.¶
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 [RFC2119] [RFC8174] when, and only when, they appear in all capitals, as shown here.¶
TODO Security¶
The definition of the EIP header as an Option for IPv6 Hop-by-hop Extension header requires the allocation of a codepoint from the "Destination Options and Hop-by-Hop Options" registry in the "Internet Protocol Version 6 (IPv6) Parameters" (https://www.iana.org/assignments/ipv6-parameters/ipv6-parameters.xhtml).¶
The definition of the EIP header as a TLV in the Segment Routing Header requires the allocation of a codepoint from the "Segment Routing Header TLVs" registry in the "Internet Protocol Version 6 (IPv6) Parameters" (https://www.iana.org/assignments/ipv6-parameters/ipv6-parameters.xhtml).¶
The definition of EIP Information Elements in the EIP header will require the definition of a IANA registry.¶
TODO acknowledge.¶