| Internet-Draft | Analysis for Multiple Data Plane Solutio | August 2024 |
| Fu, et al. | Expires 2 March 2025 | [Page] |
This document presents an overall framework for the data plane of Computing-Aware Traffic Steering (CATS). In particular, it illustrates several optional and possible data plane solutions, compares their key features and main differences, and analyzes their corresponding applicable scenarios.¶
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 2 March 2025.¶
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. 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.¶
As described in [I-D.ietf-cats-usecases-requirements], traffic steering which takes computing resource conditions and metrics into account would benefit computing-related services, including latency-sensitive services which rely upon the use of augmented reality or virtual reality (AR/VR) techniques.¶
Computing-Aware Traffic Steering (CATS) [I-D.ietf-cats-framework] aims to solve the problem that how the network edge can steer traffic between clients of a service and sites offering the service. To enable the computing- and network-aware traffic steering decisions, awareness of computing information and network information are fundamental premises. The CATS architecture is an overlay framework for the selection of the most appropriate service contact instance for placing a service request. However, the CATS framework does not assume any specific data plane and control plane solutions.¶
This document proposes several potential data plane solutions for the realization of CATS, and compares their key features and main application scenarios. These solutions use an anycast IP address or digital identification as the Computing-aware Service ID (CS-ID) associated with a service.¶
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.¶
This document makes use of the terms defined in [I-D.ietf-cats-framework].¶
As illustrated in Figure 1, underlay network infrastructure and devices are deployed between an ingress CATS-Router and service contact instances, where corresponding CATS functionality is deployed at the ingress CATS-Router 1 and egress CATS-Router 2/3. An egress CAT-router connects to multiple service sites. At a specific service site, single service contact instance or multiple service contact instances are deployed.¶
CATS overlay encapsulation is established from the ingress CATS-Router to the egress CATS-Router connected to the service site. For ease of description in this document, it is assumed that a specific tunnel between CATS-Routers is an SRv6 Policy[RFC8986].¶
+--------+ | Client | <--------------------------- +----+---+ | Phase 1 +--------+-------+ <----------------------- | C-TC | +---------+------+ | | C-PS | | +------+ +--------------+ CATS-Router 1 +--------------+ | +----------------+ | | Underlay Infrastructure | Phase 2 | SRv6 Encap | | +----------------+ +----------------+ | +-+ CATS-Router 2 +--------+ CATS-Router 3 +-+ +----------------+ +----------------+ | C-SMA | | C-SMA | +--------+-------+ +--------+-------+ <---------- | | | | +----+-----+ +---+------+ Phase 3 +--+--------+ | +--+--------+ | | Service | | | Service | | | instance +-+ | instance +-+ +-----------+ <------------+-----------+ <------------- Edge site1 Edge site2
Control plane: The ingress CATS-Router ("CATS-Router 1") receives service routes from the egress CATS-Routers ("CATS-Router 2/3") , including network and computing indicators. The C-PS determines an associated egress CATS-Router by selecting the most appropriate service site and corresponding network forwarding path based on routing strategies and policies, utilizing collected network and computing metrics. The ingress CATS-Router generates the SRv6 tunnel encapsulation from itself to the egress CATS router based on a calculated and matched SR policy.¶
Data plane: From the client to the service contact instance, packet processing and handling procedures are generally divided into at least the following successive three phases:¶
+----+ +------+ +-------------------+ +-------------------+ |IP1 +-+ |Client| |Ingress CATS-Router| |Engress CATS-Router| +-+ IP2| +------+ +-------------------+ +-------------------+ +----+ | | | | | | +---------------+ | | | | | DATA1 | | | | | +---------------+ | | | | | inner header | | | | | | +-----------+ | | | | | | |SA=clientIP| | | | | +-------------+ | | +-----------+ | | | | | DATA1 | | | | DA=Sid1 | | | | | +-------------+ | | +-----------+ | | | | | inner header| | +---------------+ | | | |+-----------+| | | outer header | | +-------------+ | | ||SA=clientIP|| | | +-----------+ | | | DATA1 | | | |+-----------+| | | | SRH | | | +-------------+ | | || DA=Sid1 || | | +-----------+ | | | inner header| | | |+-----------+| | | | SA=C-R-I | | | |+-----------+| | | +-------------+ | | +-----------+ | | ||SA=clientIP|| | | | | | DA=C-R-E | | | |+-----------+| | | | | +-----------+ | | || DA=IP1 || | | | +---------------+ | |+-----------+| | +---------------->| | +-------------+ | | Phase 1 +--------------------->| | | | Phase 2 +---------------->| | | | Phase 3 | | | | |
Figure 2 illustrates successive phases of workflow in Solution 1¶
Anycast IP is used as the destination address of the end-to-end session. Since the destination address of the user packet is translated to the IP address of the service contact instance in phase III. After the service contact instance receives the packet, the service contact instance correspondingly utilizes the incoming source address as the destination address of the response packet, and uses the IP address of the service contact instance as the source address. To eliminate influences on the host protocol stack for service contact and session establishment, the source address of the response packet must be translated to the corresponding upstream destination address, for which an SNAT process should be performed for the response packets in a downstream workflow.¶
Specifically, the NAT (Network Address Translation) function can be provided by either the egress CATS-Router or the ingress CATS-Router. In the case of the egress CATS-Router, a special SID (Segment ID) needs to be extended to indicate the NAT translation. However, for the ingress CATS-Router, no such special SID is required; it can determine the need for NAT based on the context. This allows for flexibility in choosing different solutions based on actual circumstances.¶
Among up-to-date application scenarios, some newly introduced transport protocols would support the change of connecting IP address without interrupting the traffic flows and disconnecting the connection session. In these cases, the CATS-Routers would modify the destination address of the corresponding packets to the IP address of the selected service contact instance when the client sends its upstream packet, and establish a connection through the downstream response packet from the service instance even the IP address is modified. Corresponding capable transport protocols are outside the scope of this document.¶
+----+ +------+ +-------------------+ +-------------------+ |IP1 +-+ |Client| |Ingress CATS-Router| |Engress CATS-Router| +-+ IP2| +------+ +-------------------+ +-------------------+ +----+ | | | | | | +---------------+ | | | | | DATA1 | | | | | +---------------+ | | | | | inner header | | | | | | +-----------+ | | | | | | |SA=clientIP| | | | | +-------------+ | | +-----------+ | | | | | DATA1 | | | | DA=Sid1 | | | | | +-------------+ | | +-----------+ | | | | | inner header| | +---------------+ | | | |+-----------+| | | outer header | | +-------------+ | | ||SA=clientIP|| | | +-----------+ | | | DATA1 | | | |+-----------+| | | | SRH | | | +-------------+ | | || DA=Sid1 || | | +-----------+ | | | inner header| | | |+-----------+| | | | SA=C-R-I | | | |+-----------+| | | +-------------+ | | +-----------+ | | ||SA=clientIP|| | | | | | DA=C-R-E | | | |+-----------+| | | | | +-----------+ | | || DA=IP1 || | | | +---------------+ | |+-----------+| | +---------------->| | +-------------+ | | Phase 1 +--------------------->| | | | Phase 2 +---------------->| | | | Phase 3 | | | | |
Figure 3 illustrates successive phases of workflow in Solution 2¶
In order to adapt to the destination address modification on the terminal and service host side, the protocol stack should be correspondingly modified and upgraded. As a result, downstream packets do not need SNAT translation. In the above condition, there is only unidirectional address translation process in Solution 2. In this case, the CATS-Router does not even need to maintain and administer session status for traffic flows including unidirectional translation entries, etc.¶
+----+ +------+ +-------------------+ +-------------------+ |IP1 +-+ |Client| |Ingress CATS-Router| |Engress CATS-Router| +-+--+2| +--+---+ +----------+--------+ +--------+----------+ +--+-+ | | | | | | | Option 1: | | | +---------------+ | +---------------+ | | | | DATA1 | | | DATA1 | | | | +---------------+ | +---------------+ | | | | inner header | | | inner header | | | | | +-----------+ | | | +-----------+ | | | | | |SA=clientIP| | | | |SA=clientIP| | | | +---------------+ | | +-----------+ | | | +-----------+ | | | | DATA1 | | | | DA=Sid1 | | | | | DA=Sid1 | | | | +---------------+ | | +-----------+ | | | +-----------+ | | | | inner header | | +---------------+ | +---------------+ | | | +-----------+ | | | outer header | | | outer header | | | | |SA=clientIP| | | | +-----------+ | | | +-----------+ | | | | +-----------+ | | | | SRH | | | | | SRH | | | | | | DA=Sid1 | | | | +-----------+ | | | +-----------+ | | | | +-----------+ | | | | SA=C-R-I | | | | | SA=C-R-E | | | | +---------------+ | | +-----------+ | | | +-----------+ | | | | | | DA=C-R-E | | | | | DA=IP1 | | | | | | +-----------+ | | | +-----------+ | | | | +---------------+ | +---------------+ | | | | | | | | Option 2: | | | | +--------------+ | | | | | DATA1 | | | | | +--------------+ | | | | | inner header | | | | | |+------------+| | | | | ||SA=clientIP || | | | | |+------------+| | | | | || DA=Sid1 || | | | | |+------------+| | | | | ||DMAC=IP1-MAC|| | | | | |+------------+| | | | | +--------------+ | +------------------>| | | | Phase 1 +------------------>| | | | Phase 2 +--------------------->| | | | Phase 3 | | | | |
Figure 4 illustrates successive phases of workflow in Solution 3¶
It should be noted that the client and service host stacks of this solution are not modified, and there is no IP address translation process in the above solution. However, the service instance needs to support a mentioned tunneling model, and some protocol stacks might not support the tunneling functionality.¶
| Solution 1 | Solution 2 | Solution 3 | |
|---|---|---|---|
| CATS router requirement | High | Middle | Middle |
| Client requirement | None | Middle | None |
| Service host requirement | None | Middle | Low |
| Forwarding Performance | Low | High | High |
As illustrated in Table 1, different solutions have disparate requirements for clients, service hosts, and CATS-Routers, ultimately resulting in different forwarding performances. Generally, solution 1 has the lowest requirements for terminals and service hosts, yet its forwarding performance may be the worst. Solution 2 has the most strict requirements for terminals and service hosts. In most cases, protocol stack modification is applicable to new host protocol stacks. Solution 3 requires the service host's anycast IP to be configured and deployed, and a protocol stack to support tunnels. Both solution 2 and solution 3 can provide satisfying forwarding performance. Additionally, in solution 2 and 3, CATS-Routers are not required to support the full and standard functionality of NAT.¶
To be added.¶
TBD.¶
To be added upon contributions, comments and suggestions.¶