Internet-Draft | Yang for Path Computation | July 2023 |
Busi, et al. | Expires 8 January 2024 | [Page] |
- Workgroup:
- TEAS Working Group
- Internet-Draft:
- draft-ietf-teas-yang-path-computation-21
- Published:
- Intended Status:
- Standards Track
- Expires:
A YANG Data Model for requesting path computation
Abstract
There are scenarios, typically in a hierarchical Software-Defined Networking (SDN) context, where the topology information provided by a Traffic Engineering (TE) network provider may be insufficient for its client to perform multi-domain path computation. In these cases the client would need to request the TE network provider to compute some intra-domain paths to be used by the client to choose the optimal multi-domain paths.¶
This document provides a mechanism to request path computation by augmenting the Remote Procedure Calls (RPCs) defined in RFC YYYY.¶
[RFC EDITOR NOTE: Please replace RFC YYYY with the RFC number of draft-ietf-teas-yang-te once it has been published.¶
Moreover, this document describes some use cases where the path computation request, via YANG-based protocols (e.g., NETCONF or RESTCONF), can be needed.¶
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 8 January 2024.¶
Copyright Notice
Copyright (c) 2023 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.¶
1. Introduction
There are scenarios, typically in a hierarchical Software-Defined Networking (SDN) context, where the topology information provided by a Traffic Engineering (TE) network provider may be insufficient for its client to perform multi-domain path computation. In these cases the client would need to request the TE network provider to compute some intra-domain paths that could be used together with its topology information to compute the multi-domain path.¶
These types of scenarios can be applied to different interfaces in different reference architectures:¶
- Application-Based Network Operations (ABNO) control interface [RFC7491], in which an Application Service Coordinator can request the ABNO controller to take in charge path calculation (see Figure 1 in [RFC7491]).¶
-
Abstraction and Control of TE Networks (ACTN) [RFC8453], where a controller hierarchy is defined. In the ACTN context, path computation is needed on the interface between Customer Network Controller (CNC) and Multi-Domain Service Coordinator (MDSC), called CNC-MDSC Interface (CMI), and on the interface between MSDC and Provisioning Network Controller (PNC), called MDSC-PNC Interface (MPI). [RFC8454] describes an information model for the Path Computation request.¶
Multiple protocol solutions can be used for communication between different controller hierarchical levels. This document assumes that the controllers are communicating using YANG-based protocols (e.g., NETCONF [RFC6241] or RESTCONF [RFC8040]).¶
Path Computation Elements (PCEs), controllers and orchestrators perform their operations based on Traffic Engineering Databases (TED). Such TEDs can be described, in a technology agnostic way, with the YANG data model for TE Topologies [RFC8795]. Furthermore, the technology specific details of the TED are modelled in the technology specific topology models, e.g., the [I-D.ietf-ccamp-otn-topo-yang] for Optical Transport Network (OTN) Optical Data Unit (ODU) technologies, which augment the common TE topology model in [RFC8795].¶
The availability of such topology models allows the provisioning of the TED using YANG-based protocols (e.g., NETCONF or RESTCONF). Furthermore, it enables a PCE/controller performing the necessary abstractions or modifications and offering this customized topology to another PCE/controller or high level orchestrator.¶
The tunnels that can be provided over the networks described with the topology models can be also set-up, deleted and modified via YANG- based protocols (e.g., NETCONF or RESTCONF) using the TE tunnel YANG data model [I-D.ietf-teas-yang-te].¶
This document defines a YANG data model [RFC7950] that augments the RPC defined in [I-D.ietf-teas-yang-te]. The use of this RPC is complimentary to the configuration of a TE tunnel path in "compute-only" mode, as described in [I-D.ietf-teas-yang-te].¶
The YANG data model definition does not make any assumption about whether that the client or the server implement a "PCE" functionality, as defined in [RFC4655], and the Path Computation Element Communication Protocol (PCEP) protocol, as defined in [RFC5440].¶
Moreover, this document describes some use cases where a path computation request, via YANG-based protocols (e.g., NETCONF or RESTCONF), can be needed.¶
The YANG data model defined in this document conforms to the Network Management Datastore Architecture [RFC8342].¶
1.1. Terminology
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.¶
TED:¶
-
The traffic engineering database is a collection of all TE information about all TE nodes and TE links in a given network.¶
PCE:¶
-
A Path Computation Element (PCE) is an entity that is capable of computing a network path or route based on a network graph, and of applying computational constraints during the computation. The PCE entity is an application that can be located within a network node or component, on an out-of-network server, etc. For example, a PCE would be able to compute the path of a TE Label Switched Path (LSP) by operating on the TED and considering bandwidth and other constraints applicable to the TE LSP service request. [RFC4655].¶
Domain:¶
-
TE information is the data relating to nodes and TE links that is used in the process of selecting a TE path. TE information is usually only available within a network. We call such a zone of visibility of TE information a domain. An example of a domain may be an IGP area or an Autonomous System. [RFC7926]¶
The terminology for describing YANG data models is found in [RFC7950].¶
1.2. Tree Diagram
Tree diagrams used in this document follow the notation defined in [RFC8340].¶
1.3. Prefixes in Data Node Names
In this document, names of data nodes and other data model objects are prefixed using the standard prefix associated with the corresponding YANG imported modules, as shown in Table 1.¶
Prefix | YANG module | Reference |
---|---|---|
te-types | ietf-te-types | [RFCZZZZ] |
te | ietf-te | [RFCYYYY] |
te-pc | ietf-te-path-computation | RFCXXXX |
RFC Editor Note: Please replace XXXX with the RFC number assigned to this document. Please replace YYYY with the RFC number of [I-D.ietf-teas-yang-te] once it has been published. Please replace ZZZZ with the RFC number of [I-D.ietf-teas-rfc8776-update] once it has been published. Please remove this note.¶
2. Use Cases
This section presents some use cases, where a client needs to request underlying SDN controllers for path computation.¶
The use of the YANG data model defined in this document is not restricted to these use cases but can be used in any other use case when deemed useful.¶
The presented uses cases have been grouped, depending on the different underlying topologies: Packet/Optical Integration (Section 2.1); multi-domain Traffic Engineered (TE) Networks (Section 2.2); and Data Center Interconnections (Section 2.3). Use cases in Section 2.4 and Section 2.5 respectively present how to apply this YANG data model for standard multi-domain PCE i.e. Backward Recursive Path Computation [RFC5441] and Hierarchical PCE [RFC6805].¶
2.1. Packet/Optical Integration
In this use case, an optical domain is used to provide connectivity to some nodes of a packet domain (see Figure 1).¶
Figure 1 as well as Figure 2 describe two different examples of packet/optical topologies and only show a partial view of the packet network connectivity, before additional packet connectivity is provided by the optical network.¶
It is assumed that the Optical Domain Controller provides to the Packet/Optical Coordinator an abstracted view of the optical network. A possible abstraction could be to represent the whole optical network as one "virtual node" with "virtual ports" connected to the access links, as shown in Figure 2.¶
It is also assumed that Packet Domain Controller can provide the Packet/Optical Coordinator the information it needs to set up connectivity between packet nodes through the optical network (e.g., the access links).¶
The path computation request helps the Packet/Optical Coordinator to know the real connections that can be provided by the optical network.¶
In this use case, the Packet/Optical Coordinator needs to set up an optimal underlying path for an IP link between R1 and R2.¶
As depicted in Figure 2, the Packet/Optical Coordinator has only an "abstracted view" of the physical network, and it does not know the feasibility or the cost of the possible optical paths (e.g., VP1-VP4 and VP2-VP5), which depend on the current status of the physical resources within the optical network and on vendor-specific optical attributes.¶
The Packet/Optical Coordinator can request the underlying Optical Domain Controller to compute a set of potential optimal paths, taking into account optical constraints. Then, based on its own constraints, policy and knowledge (e.g. cost of the access links), it can choose which one of these potential paths to use to set up the optimal end- to-end path crossing optical network.¶
For example, in Figure 3, the Packet/Optical Coordinator can request the Optical Domain Controller to compute the paths between VP1-VP4 and VP2-VP5 and then decide to set up the optimal end-to-end path using the VP2-VP5 optical path even if this is not the optimal path from the optical domain perspective.¶
Considering the dynamicity of the connectivity constraints of an optical domain, it is possible that a path computed by the Optical Domain Controller when requested by the Packet/Optical Coordinator is no longer valid/available when the Packet/Optical Coordinator requests it to be set up. This is further discussed in Section 3.3.¶
2.2. Multi-domain TE networks
In this use case there are two TE domains which are interconnected together by multiple inter-domains links.¶
A possible example could be a multi-domain optical network.¶
In order to set up an end-to-end multi-domain TE path (e.g., between nodes A and H), the Multi-Domain Controller needs to know the feasibility or the cost of the possible TE paths within the two TE domains, which depend on the current status of the physical resources within each TE domain. This is more challenging in case of optical networks because the optimal paths depend also on vendor-specific optical attributes (which may be different in the two domains if they are provided by different vendors).¶
In order to set up a multi-domain TE path (e.g., between nodes A and H), the Multi-Domain Controller can request the TE Domain Controllers to compute a set of intra-domain optimal paths and take decisions based on the information received. For example:¶
- The Multi-Domain Controller asks TE Domain Controllers to provide set of paths between A-C, A-D, E-H and F-H¶
- TE Domain Controllers return a set of feasible paths with the associated costs: the path A-C is not part of this set (in optical networks, it is typical to have some paths not being feasible due to optical constraints that are known only by the Optical Domain Controller)¶
- The Multi-Domain Controller will select the path A-D-F-H since it is the only feasible multi-domain path and then request the TE Domain Controllers to set up the A-D and F-H intra-domain paths¶
-
If there are multiple feasible paths, the Multi-Domain Controller can select the optimal path knowing the cost of the intra-domain paths (provided by the TE domain controllers) and the cost of the inter-domain links (known by the Multi-Domain Controller)¶
This approach may have some scalability issues when the number of TE domains is quite big (e.g. 20).¶
In this case, it would be worthwhile using the abstract TE topology information provided by the TE Domain Controllers to limit the number of potential optimal end-to-end paths and then request path computation from fewer TE Domain Controllers in order to decide what the optimal path within this limited set is.¶
For more details, see Section 3.2.3.¶
2.3. Data Center Interconnections
In these use case, there is a TE domain which is used to provide connectivity between Data Centers which are connected with the TE domain using access links.¶
In this use case, there is the need to transfer data from Data Center 1 (DC1) to either DC2 or DC3 (e.g. workload migration).¶
The optimal decision depends both on the cost of the TE path (DC1-DC2 or DC1-DC3) and of the Data Center resources within DC2 or DC3.¶
The Cloud Network Orchestrator needs to make a decision for optimal connection based on TE network constraints and Data Center resources.¶
It may not be able to make this decision because it has only an abstract view of the TE network (as in Section 2.1).¶
The Cloud Network Orchestrator can request to the TE Network Controller to compute the cost of the possible TE paths (e.g., DC1- DC2 and DC1-DC3) and to the DC Controller to provide the information it needs about the required Data Center resources within DC2 and DC3 and then it can take the decision about the optimal solution based on this information and its policy.¶
2.4. Backward Recursive Path Computation scenario
[RFC5441] has defined the Virtual Source Path Tree (VSPT) flag within the RP (Request Parameters) object in order to compute inter-domain paths following a "Backward Recursive Path Computation" (BRPC) method. The main principle is to forward the PCReq message up to the destination domain. Then, each PCE involved in the computation will compute its part of the path and send it back to the requester through PCE Response message. The resulting computation is spread from destination PCE to source PCE. Each PCE is in charge of merging the path it received with the one it calculated. At the end, the source PCE merges its local part of the path with the received one to achieve the end-to-end path.¶
Figure 6 below show a typical BRPC scenario where 3 PCEs cooperate to compute inter-domain paths.¶
In this use case, a client can use the YANG data model defined in this document to request path computation from the PCE that controls the source of the tunnel. For example, a client can request to the PCE of domain A to compute a path from a source S, within domain A, to a destination D, within domain C. Then PCE of domain A will use PCEP protocol, as per [RFC5441], to compute the path from S to D and in turn gives the final answer to the requester.¶
2.5. Hierarchical PCE scenario
[RFC6805] has defined an architecture and extensions to the PCE standard to compute inter-domain path following a hierarchical method. Two new roles have been defined: parent PCE and child PCE. The parent PCE is in charge to coordinate the end-to-end path computation. For that purpose it sends to each child PCE involved in the multi-domain path computation a PCE Request message to obtain the local part of the path. Once received all answer through PCE Response message, the parent PCE will merge the different local parts of the path to achieve the end-to-end path.¶
Figure 7 below shows a typical hierarchical scenario where a parent PCE request end-to-end path to the different child PCE. Note that a PCE could take independently the role of child or parent PCE depending of which PCE will request the path.¶
In this use case, a client can use the YANG data model defined in this document to request to the parent PCE a path from a source S to a destination D. The parent PCE will in turn contact the child PCEs through PCEP protocol to compute the end-to-end path and then return the computed path to the client, using the YANG data model defined in this document. For example the YANG data model can be used to request to PCE5 acting as parent PCE to compute a path from source S, within domain 1, to destination D, within domain 3. PCE5 will contact child PCEs of domain 1, 2 and 3 to obtain local part of the end-to-end path through the PCEP protocol. Once received the PCRep message, it merges the answers to compute the end-to-end path and send it back to the client.¶
3. Motivations
This section provides the motivation for the YANG data model defined in this document.¶
Section 3.1 describes the motivation for a YANG data model to request path computation.¶
Section 3.2 describes the motivation for a YANG data model which complements the TE topology YANG data model defined in [RFC8795].¶
Section 3.3 describes the motivation for a YANG RPC which complements the TE tunnel YANG data model defined in [I-D.ietf-teas-yang-te].¶
3.1. Motivation for a YANG Model
3.1.1. Benefits of common data models
The YANG data model for requesting path computation is closely aligned with the YANG data models that provide (abstract) TE topology information, i.e., [RFC8795] as well as that are used to configure and manage TE tunnels, i.e., [I-D.ietf-teas-yang-te].¶
There are many benefits in aligning the data model used for path computation requests with the YANG data models used for TE topology information and for TE tunnels configuration and management:¶
3.1.2. Benefits of a single interface
The system integration effort is typically lower if a single, consistent interface is used by controllers, i.e., one data modeling language (i.e., YANG) and a common protocol (e.g., NETCONF or RESTCONF).¶
Practical benefits of using a single, consistent interface include:¶
- Simple authentication and authorization: The interface between different components has to be secured. If different protocols have different security mechanisms, ensuring a common access control model may result in overhead. For instance, there may be a need to deal with different security mechanisms, e.g., different credentials or keys. This can result in increased integration effort.¶
- Consistency: Keeping data consistent over multiple different interfaces or protocols is not trivial. For instance, the sequence of actions can matter in certain use cases, or transaction semantics could be desired. While ensuring consistency within one protocol can already be challenging, it is typically cumbersome to achieve that across different protocols.¶
- Testing: System integration requires comprehensive testing, including corner cases. The more different technologies are involved, the more difficult it is to run comprehensive test cases and ensure proper integration.¶
- Middle-box friendliness: Provider and consumer of path computation requests may be located in different networks, and middle-boxes such as firewalls, NATs, or load balancers may be deployed. In such environments it is simpler to deploy a single protocol. Also, it may be easier to debug connectivity problems.¶
- Tooling reuse: Implementers may want to implement path computation requests with tools and libraries that already exist in controllers and/or orchestrators, e.g., leveraging the rapidly growing eco-system for YANG tooling.¶
3.1.3. Extensibility
Path computation is only a subset of the typical functionality of a controller. In many use cases, issuing path computation requests comes along with the need to access other functionality on the same system. In addition to obtaining TE topology, for instance also configuration of services (set-up/modification/deletion) may be required, as well as:¶
- Receiving notifications for topology changes as well as integration with fault management¶
- Performance management such as retrieving monitoring and telemetry data¶
- Service assurance, e.g., by triggering OAM functionality¶
-
Other fulfilment and provisioning actions beyond tunnels and services, such as changing QoS configurations¶
YANG is a very extensible and flexible data modeling language that can be used for all these use cases.¶
3.2. Interactions with TE topology
The use cases described in Section 2 have been described assuming that the topology view exported by each underlying controller to its client is aggregated using the "virtual node model", defined in [RFC7926].¶
TE topology information, e.g., as provided by [RFC8795], could in theory be used by an underlying controller to provide TE information to its client thus allowing a PCE available within its client to perform multi-domain path computation on its own, without requesting path computations to the underlying controllers.¶
In case the client does not implement a PCE function, as discussed in Section 1, it could not perform path computation based on TE topology information and would instead need to request path computation from the underlying controllers to get the information it needs to find the optimal end-to-end path.¶
In case the client implements a PCE function, as discussed in Section 1, the TE topology information needs to be complete and accurate, which would bring to scalability issues.¶
Using TE topology to provide a "virtual link model" aggregation, as described in [RFC7926], may be insufficient, unless the aggregation provides complete and accurate information, which would still cause scalability issues, as described in Section 3.2.1 below.¶
Using TE topology abstraction, as described in [RFC7926], may lead to compute an unfeasible path, as described in [RFC7926] in Section 3.2.2 below.¶
Therefore when computing an optimal multi-domain path, there is a scalability trade-off between providing complete and accurate TE information and the number of path computation requests to the underlying controllers.¶
The TE topology information used, in a complimentary way, to reduce the number for path computation requests to the underlying controllers, are described in Section 3.2.3 below.¶
3.2.1. TE topology aggregation
Using the TE topology model, as defined in [RFC8795], the underlying controller can export the whole TE domain as a single TE node with a "detailed connectivity matrix" (which provides specific TE attributes, such as delay, Shared Risk Link Groups (SRLGs) and other TE metrics, between each ingress and egress links).¶
The information provided by the "detailed connectivity matrix" would be equivalent to the information that should be provided by "virtual link model" as defined in [RFC7926].¶
For example, in the Packet/Optical Integration use case, described in Section 2.1, the Optical Domain Controller can make the information shown in Figure 3 available to the Packet/Optical Coordinator as part of the TE topology information and the Packet/Optical Coordinator could use this information to calculate on its own the optimal path between R1 and R2, without requesting any additional information to the Optical Domain Controller.¶
However, when designing the amount of information to provide within the "detailed connectivity matrix", there is a tradeoff to be considered between accuracy (i.e., providing "all" the information that might be needed by the PCE available within the client) and scalability.¶
Figure 8 below shows another example, similar to Figure 3, where there are two possible Optical paths between VP1 and VP4 with different properties (e.g., available bandwidth and cost).¶
If the information in the "detailed connectivity matrix" is not complete/accurate, we can have the following drawbacks:¶
- If only the VP1-VP4 path with available bandwidth of 2 Gb/s and cost 50 is reported, the client's PCE will fail to compute a 5 Gb/s path between routers R1 and R2, although this would be feasible;¶
-
If only the VP1-VP4 path with available bandwidth of 10 Gb/s and cost 65 is reported, the client's PCE will compute, as optimal, the 1 Gb/s path between R1 and R2 going through the VP2-VP5 path within the optical domain while the optimal path would actually be the one going thought the VP1-VP4 sub-path (with cost 50) within the optical domain.¶
Reporting all the information, as in Figure 8, using the "detailed connectivity matrix", is quite challenging from a scalability perspective. The amount of this information is not just based on number of end points (which would scale as N-square), but also on many other parameters, including client rate, user constraints/policies for the service, e.g. max latency < N ms, max cost, etc., exclusion policies to route around busy links, min Optical Signal to Noise Ratio (OSNR) margin, max pre-Forward Error Correction (FEC) Bit Error Rate (BER) etc. All these constraints could be different based on connectivity requirements.¶
It is also worth noting that the "connectivity matrix" has been originally defined in Wavelength Switched Optical Networks (WSON), [RFC7446], to report the connectivity constrains of a physical node within the Wavelength Division Multiplexing (WDM) network: the information it contains is pretty "static" and therefore, once taken and stored in the TE data base, it can be always being considered valid and up-to-date in path computation request.¶
The "connectivity matrix" is sometimes confused with "optical reach table" that contain multiple (e.g. k-shortest) regen-free reachable paths for every A-Z node combination in the network. Optical reach tables can be calculated offline, utilizing vendor optical design and planning tools, and periodically uploaded to the Controller: these optical path reach tables are fairly static. However, to get the connectivity matrix, between any two sites, either a regen free path can be used, if one is available, or multiple regen free paths are concatenated to get from the source to the destination, which can be a very large combination. Additionally, when the optical path within optical domain needs to be computed, it can result in different paths based on input objective, constraints, and network conditions. In summary, even though "optical reach table" is fairly static, which regen free paths to build the connectivity matrix between any source and destination is very dynamic, and is done using very sophisticated routing algorithms.¶
Using the "basic connectivity matrix" with an abstract node to abstract the information regarding the connectivity constraints of an Optical domain, would make this information more "dynamic" since the connectivity constraints of an optical domain can change over time because some optical paths that are feasible at a given time may become unfeasible at a later time when e.g., another optical path is established.¶
The information in the "detailed connectivity matrix" is even more dynamic since the establishment of another optical path may change some of the parameters (e.g., delay or available bandwidth) in the "detailed connectivity matrix" while not changing the feasibility of the path.¶
There is therefore the need to keep the information in the "detailed connectivity matrix" updated which means that there another tradeoff between the accuracy (i.e., providing "all" the information that might be needed by the client's PCE) and having up-to-date information. The more the information is provided and the longer it takes to keep it up-to-date which increases the likelihood that the client's PCE computes paths using not updated information.¶
It seems therefore quite challenging to have a "detailed connectivity matrix" that provides accurate, scalable and updated information to allow the client's PCE to take optimal decisions by its own.¶
Considering the example in Figure 8 with the approach defined in this document, the client, when it needs to set up an end-to-end path, it can request the Optical Domain Controller to compute a set of optimal paths (e.g., for VP1-VP4 and VP2-VP5) and take decisions based on the information received:¶
- When setting up a 5 Gb/s path between routers R1 and R2, the Optical Domain Controller may report only the VP1-VP4 path as the only feasible path: the Packet/Optical Coordinator can successfully set up the end-to-end path passing though this optical path;¶
- When setting up a 1 Gb/s path between routers R1 and R2, the Optical Domain Controller (knowing that the path requires only 1 Gb/s) can report both the VP1-VP4 path, with cost 50, and the VP2- VP5 path, with cost 65. The Packet/Optical Coordinator can then compute the optimal path which is passing thought the VP1-VP4 sub- path (with cost 50) within the optical domain.¶
3.2.2. TE topology abstraction
Using the TE topology model, as defined in [RFC8795], the underlying controller can export to its client an abstract TE topology, composed by a set of TE nodes and TE links, representing the abstract view of the topology under its control.¶
For example, in the multi-domain TE network use case, described in Section 2.2, the TE Domain Controller 1 can export a TE topology encompassing the TE nodes A, B, C and D and the TE links interconnecting them. In a similar way, the TE Domain Controller 2 can export a TE topology encompassing the TE nodes E, F, G and H and the TE links interconnecting them.¶
In this example, for simplicity reasons, each abstract TE node maps with each physical node, but this is not necessary.¶
In order to set up a multi-domain TE path (e.g., between nodes A and H), the Multi-Domain Controller can compute by its own an optimal end-to-end path based on the abstract TE topology information provided by the domain controllers. For example:¶
- Multi-Domain Controller can compute, based on its own TED data, the optimal multi-domain path being A-B-C-E-G-H, and then request the TE Domain Controllers to set up the A-B-C and E-G-H intra- domain paths¶
- But, during path set-up, the TE Domain Controller may find out that A-B-C intra-domain path is not feasible (as discussed in Section 2.2, in optical networks it is typical to have some paths not being feasible due to optical constraints that are known only by the Optical Domain Controller), while only the path A-B-D is feasible¶
-
So what the Multi-Domain Controller has computed is not good and it needs to re-start the path computation from scratch¶
As discussed in Section 3.2.1, providing more extensive abstract information from the TE Domain Controllers to the Multi-Domain Controller may lead to scalability problems.¶
In a sense this is similar to the problem of routing and wavelength assignment within an optical domain. It is possible to do first routing (step 1) and then wavelength assignment (step 2), but the chances of ending up with a good path is low. Alternatively, it is possible to do combined routing and wavelength assignment, which is known to be a more optimal and effective way for optical path set-up. Similarly, it is possible to first compute an abstract end-to-end path within the Multi-Domain Controller (step 1) and then compute an intra-domain path within each optical domain (step 2), but there are more chances not to find a path or to get a suboptimal path than by performing multiple per-domain path computations and then stitching them together.¶
3.2.3. Complementary use of the TE topology
As discussed in Section 2.2, there are some scalability issues with path computation requests in a multi-domain TE network with many TE domains, in terms of the number of requests to send to the TE Domain Controllers. It would therefore be worthwhile using the abstract TE topology information provided by the TE Domain Controllers to limit the number of requests.¶
An example can be described considering the multi-domain abstract TE topology shown in Figure 9. In this example, an end-to-end TE path between domains A and F needs to be set up. The transit TE domain should be selected between domains B, C, D and E.¶
The actual cost of each intra-domain path is not known a priori from the abstract topology information. The Multi-Domain Controller only knows, from the TE topology provided by the underlying domain controllers, the feasibility of some intra-domain paths and some upper-bound and/or lower-bound cost information. With this information, together with the cost of inter-domain links, the Multi- Domain Controller can understand by its own that:¶
- Domain B cannot be selected as the path connecting domains A and F is not feasible;¶
-
Domain E cannot be selected as a transit domain since it is known from the abstract topology information provided by domain controllers that the cost of the multi-domain path A-E-F (which is 100, in the best case) will be always be higher than the cost of the multi-domain paths A-D-F (which is 90, in the worst case) and A-C-F (which is 80, in the worst case).¶
Therefore, the Multi-Domain Controller can understand by its own that the optimal multi-domain path could be either A-D-F or A-C-F but it cannot know which one of the two possible option actually provides the optimal end-to-end path.¶
The Multi-Domain Controller can therefore request path computation only to the TE Domain Controllers A, D, C and F (and not to all the possible TE Domain Controllers).¶
Based on these requests, the Multi-Domain Controller can know the actual cost of each intra-domain paths which belongs to potential optimal end-to-end paths, as shown in Figure 10, and then compute the optimal end-to-end path (e.g., A-D-F, having total cost of 50, instead of A-C-F having a total cost of 70).¶
3.3. Path Computation RPC
The TE tunnel YANG data model, defined in [I-D.ietf-teas-yang-te], can support the need to request path computation, as described in section 5.1.2 of [I-D.ietf-teas-yang-te].¶
This solution is stateful since the state of each created "compute- only" TE tunnel path needs to be maintained, in the YANG datastores (at least in the running datastore and operational datastore), and updated, when underlying network conditions change.¶
The RPC mechanism allows requesting path computation using a simple atomic operation, without creating any state in the YANG datastores, and it is the natural option/choice, especially with stateless PCE.¶
It is very useful to provide both the options of using an RPC as well as of setting up TE tunnel paths in "compute-only" mode. It is suggested to use the RPC as much as possible and to rely on "compute-only" TE tunnel paths, when really needed.¶
Using the RPC solution would imply that the underlying controller (e.g., a PNC) computes a path twice during the process to set up an LSP: at time T1, when its client (e.g., an MDSC) sends a path computation RPC request to it, and later, at time T2, when the same client (MDSC) creates a TE tunnel requesting the set-up of the LSP. The underlying assumption is that, if network conditions have not changed, the same path that has been computed at time T1 is also computed at time T2 by the underlying controller (e.g. PNC) and therefore the path that is set up at time T2 is exactly the same path that has been computed at time T1.¶
However, since the operation is stateless, there is no guarantee that the returned path would still be available when path set-up is requested: this does not cause major issues when the time between path computation and path set-up is short (especially if compared with the time that would be needed to update the information of a very detailed connectivity matrix).¶
In most of the cases, there is even no need to guarantee that the path that has been set up is the exactly same as the path that has been returned by path computation, especially if it has the same or even better metrics. Depending on the abstraction level applied by the server, the client may also not know the actual computed path.¶
The most important requirement is that the required global objectives (e.g., multi-domain path metrics and constraints) are met. For this reason a path verification phase is always necessary to verify that the actual path that has been set up meets the global objectives (for example in a multi-domain network, the resulting end-to-end path meets the required end-to-end metrics and constraints).¶
In most of the cases, even if the path being set up is not exactly the same as the path returned by path computation, its metrics and constraints are "good enough" and the path verification passes successfully. In the few corner cases where the path verification fails, it is possible repeat the whole process (path computation, path set-up and path verification).¶
In case it is required to set up at T2 exactly the same path computed at T1, the RPC solution should not be used and, instead, a "compute- only" TE tunnel path should be set up, allowing also notifications in case the computed path has been changed.¶
In this case, at time T1, the client (MDSC) creates a TE tunnel in a compute-only mode in the running datastore and later, at time T2, changes the configuration of that TE tunnel (not to be any more in a compute-only mode) to trigger the set-up of the LSP over the path which have been computed at time T1 and reported in the operational datastore.¶
It is worth noting that also using the "compute-only" TE tunnel path, although increasing the likelihood that the computed path is available at path set-up, does not guarantee that because notifications may not be reliable or delivered on time. Path verification is needed also in this case.¶
The solution based on "compute-only" TE tunnel path has also the following drawbacks:¶
- Several messages required for any path computation¶
- Requires persistent storage in the underlying controller¶
- Need for garbage collection for stranded paths¶
- Process burden to detect changes on the computed paths in order to provide notifications update¶
3.3.1. Temporary reporting of the computed path state
This section describes an optional extension to the stateless behavior of the path computation RPC, where the underlying controller, after having received a path computation RPC request, maintains some "transient state" associated with the computed path, allowing the client to request the set-up of exactly that path, if still available.¶
This is similar to the "compute-only" TE tunnel path solution but, to avoid the drawbacks of the stateful approach, is leveraging the path computation RPC and the separation between configuration and operational datastore, as defined in the NMDA architecture [RFC8342].¶
The underlying controller, after having computed a path, as requested by a path computation RPC, also creates a TE tunnel instance within the operational datastore, to store that computed path. This would be similar to a "compute-only" TE tunnel path, with the only difference that there is no associated TE tunnel instance within the running datastore.¶
Since the underlying controller stores in the operational datastore the computed path based on an abstract topology it exposes, it also remembers, internally, which is the actual native path (physical path), within its native topology (physical topology), associated with that compute-only TE tunnel instance.¶
Afterwards, the client (e.g., MDSC) can request the set-up of that specific path by creating a TE tunnel instance (not in compute-only mode) in the running datastore using the same tunnel-name of the existing TE tunnel in the operational datastore: this will trigger the underlying controller to set up that path, if still available.¶
There are still cases where the path being set up is not exactly the same as the path that has been computed:¶
- When the tunnel is configured with path constraints which are not compatible with the computed path;¶
- When the tunnel set-up is requested after the resources of the computed path are no longer available;¶
-
When the tunnel set-up is requested after the computed path is no longer known (e.g. due to a server reboot) by the underlying controller.¶
In all these cases, the underlying controller should compute and set up a new path.¶
Therefore the "path verification" phase, as described in Section 3.3 above, is always needed to check that the path that has been set up is still "good enough".¶
Since this new approach is not completely stateless, garbage collection is implemented using a timeout that, when it expires, triggers the removal of the computed path from the operational datastore. This operation is fully controlled by the underlying controller without the need for any action to be taken by the client that is not able to act on the operational datastore. The default value of this timeout is 10 minutes but a different value may be configured by the client.¶
In addition, it is possible for the client to tag each path computation request with a transaction-id allowing for a faster removal of all the paths associated with a transaction-id, without waiting for their timers to expire.¶
The underlying controller can remove from the operational datastore all the paths computed with a given transaction-id which have not been set up either when it receives a Path Delete RPC request for that transaction-id or, automatically, right after the set-up up of a path that has been previously computed with that transaction-id.¶
This possibility is useful when multiple paths are computed but, at most, only one is set up (e.g., in multi-domain path computation scenario scenarios). After the selected path has been set up (e.g, in one domain during multi-domain path set-up), all the other alternative computed paths can be automatically deleted by the underlying controller (since no longer needed). The client can also request, using the Path Delete RPC request, the underlying controller to remove all the computed paths, if none of them is going to be set up (e.g., in a transit domain not being selected by multi-domain path computation and so not being automatically deleted).¶
This approach is complimentary and not alternative to the timer which is always needed to avoid stranded computed paths being stored in the operational datastore when no path is set up and no explicit Path Delete RPC request is received.¶
4. Path computation and optimization for multiple paths
There are use cases, where it is advantageous to request path computation for a set of paths, through a network or through a network domain, using a single request [RFC5440].¶
In this case, sending a single request for multiple path computations, instead of sending multiple requests for each path computation, would reduce the protocol overhead and it would consume less resources (e.g., threads in the client and server).¶
In the context of a typical multi-domain TE network, there could multiple choices for the ingress/egress points of a domain and the Multi-Domain Controller needs to request path computation between all the ingress/egress pairs to select the best pair. For example, in the example of Section 2.2, the Multi-Domain Controller needs to request the TE Network Controller 1 to compute the A-C and the A-D paths and to the TE Network Controller 2 to compute the E-H and the F-H paths.¶
It is also possible that the Multi-Domain Controller receives a request to set up a group of multiple end to end connections. The Multi-Domain Controller needs to request each TE domain Controller to compute multiple paths, one (or more) for each end to end connection.¶
There are also scenarios where it can be needed to request path computation for a set of paths in a synchronized fashion.¶
One example could be computing multiple diverse paths. Computing a set of diverse paths in an unsynchronized fashion, leads to the possibility of not being able to satisfy the diversity requirement. In this case, it is preferable to compute a sub-optimal primary path for which a diversely routed secondary path exists.¶
There are also scenarios where it is needed to request optimizing a set of paths using objective functions that apply to the whole set of paths, see [RFC5541], e.g. to minimize the sum of the costs of all the computed paths in the set.¶
5. YANG data model for requesting Path Computation
This document define a YANG RPC to request path computation as an "augmentation" of tunnel-rpc, defined in [I-D.ietf-teas-yang-te]. This model provides the RPC input attributes that are needed to request path computation and the RPC output attributes that are needed to report the computed paths.¶
augment /te:tunnels-path-compute/te:input/te:path-compute-info: +-- path-request* [request-id] | ... augment /te:tunnels-path-compute/te:output/te:path-compute-result: +--ro response* [response-id] +--ro response-id uint32 +--ro computed-paths-properties | +--ro computed-path-properties* [k-index] | ...¶
This model extensively re-uses the grouping defined in [I-D.ietf-teas-yang-te] and [I-D.ietf-teas-rfc8776-update] to ensure maximal syntax and semantics commonality.¶
This YANG data model allows one RPC to include multiple path requests, each path request being identified by a request-id. Therefore, one RPC can return multiple responses, one for each path request, being identified by a response-id equal to the corresponding request-id. Each response reports one or more computed paths, as requested by the k-requested-paths attribute. By default, each response reports one computed path.¶
5.1. Synchronization of multiple path computation requests
The YANG data model permits the synchronization of a set of multiple path requests (identified by specific request-id) all related to a "svec" container emulating the syntax of the Synchronization VECtor (SVEC) PCEP object, defined in [RFC5440].¶
+-- synchronization* [] +-- svec | +-- relaxable? boolean | +-- disjointness? te-path-disjointness | +-- request-id* uint32 +-- svec-constraints | +-- path-metric-bound* [metric-type] | +-- metric-type identityref | +-- upper-bound? uint64 +-- path-srlgs-lists | +-- path-srlgs-list* [usage] | +-- usage identityref | +-- values* srlg +-- path-srlgs-names | +-- path-srlgs-name* [usage] | +-- usage identityref | +-- names* string +-- exclude-objects | +-- excludes* [] | +-- (type)? | +--:(numbered-node-hop) | | +-- numbered-node-hop | | +-- node-id te-node-id | | +-- hop-type? te-hop-type | +--:(numbered-link-hop) | | +-- numbered-link-hop | | +-- link-tp-id te-tp-id | | +-- hop-type? te-hop-type | | +-- direction? te-link-direction | +--:(unnumbered-link-hop) | | +-- unnumbered-link-hop | | +-- link-tp-id te-tp-id | | +-- node-id te-node-id | | +-- hop-type? te-hop-type | | +-- direction? te-link-direction | +--:(as-number) | | +-- as-number-hop | | +-- as-number inet:as-number | | +-- hop-type? te-hop-type | +--:(label) | +-- label-hop | +-- te-label | ... +-- optimizations +-- (algorithm)? +--:(metric) {te-types:path-optimization-metric}? | +-- optimization-metric* [metric-type] | +-- metric-type identityref | +-- weight? uint8 +--:(objective-function) {te-types:path-optimization-objective-function}? +-- objective-function +-- objective-function-type? identityref¶
The model, in addition to the metric types, defined in [I-D.ietf-teas-rfc8776-update], which can be applied to each individual path request, supports also additional metric types, which apply to a set of synchronized requests, as referenced in [RFC5541]. These additional metric types are defined by the following YANG identities:¶
- svec-metric-type: base YANG identity from which cumulative metric types identities are derived.¶
- svec-metric-cumul-te: cumulative TE cost metric type, as defined in [RFC5541].¶
- svec-metric-cumul-igp: cumulative IGP cost metric type, as defined in [RFC5541].¶
- svec-metric-cumul-hop: cumulative Hop metric type, representing the cumulative version of the Hop metric type defined in [I-D.ietf-teas-rfc8776-update].¶
- svec-metric-aggregate-bandwidth-consumption: aggregate bandwidth consumption metric type, as defined in [RFC5541].¶
- svec-metric-load-of-the-most-loaded-link: load of the most loaded link metric type, as defined in [RFC5541].¶
5.2. Returned metric values
This YANG data model provides a way to return the values of the metrics computed by the path computation in the output of RPC, together with other important information (e.g. SRLG, affinities, explicit route), emulating the syntax of the "C" flag of the "METRIC" PCEP object [RFC5440]:¶
| +--ro path-properties | +--ro path-metric* [metric-type] | | +--ro metric-type identityref | | +--ro accumulative-value? uint64 | +--ro path-affinities-values | | +--ro path-affinities-value* [usage] | | +--ro usage identityref | | +--ro value? admin-groups | +--ro path-affinity-names | | +--ro path-affinity-name* [usage] | | +--ro usage identityref | | +--ro affinity-name* [name] | | +--ro name string | +--ro path-srlgs-lists | | +--ro path-srlgs-list* [usage] | | +--ro usage identityref | | +--ro values* srlg | +--ro path-srlgs-names | | +--ro path-srlgs-name* [usage] | | +--ro usage identityref | | +--ro names* string | +--ro path-route-objects | | ... | +--ro te-bandwidth | | ... | +--ro disjointness-type? | te-types:te-path-disjointness¶
It also allows the client to request which information (metrics, SRLG and/or affinities) should be returned:¶
| +-- request-id uint32 | ... | +-- requested-metrics* [metric-type] | | +-- metric-type identityref | +-- return-srlgs? boolean | +-- return-affinities? boolean | ...¶
This feature is essential for path computation in a multi-domain TE network as described in Section 2.2. In this case, the metrics returned by a path computation requested to a given underlying controller must be used by the client to compute the best end-to-end path. If they are missing, the client cannot compare different paths calculated by the underlying controllers and choose the best one for the optimal end-to-end (e2e) path.¶
5.3. Multiple Paths Requests for the same TE tunnel
The YANG data model allows including multiple requests for different paths intended to be used within the same tunnel or within different tunnels.¶
When multiple requested paths are intended to be used within the same tunnel (e.g., requesting path computation for the primary and secondary paths of a protected tunnel), the set of attributes that are intended to be configured on per-tunnel basis rather than on per- path basis are common to all these path requests. These attributes includes both attributes which can be configured only a per-tunnel basis (e.g., tunnel-name, source/destination TTP, encoding and switching-type) as well attributes which can be configured both on a per-tunnel and on a per-path basis (e.g., the te-bandwidth or the associations).¶
Therefore, a tunnel-attributes list is defined, within the path computation request RPC:¶
+-- tunnel-attributes* [tunnel-name] | +-- tunnel-name string | +-- encoding? identityref | +-- switching-type? identityref | +-- source? te-types:te-node-id | +-- destination? te-types:te-node-id | +-- src-tunnel-tp-id? binary | +-- dst-tunnel-tp-id? binary | +-- bidirectional? boolean | +-- association-objects | | ... | +-- protection-type? identityref | +-- restoration-type? identityref | +-- restoration-scheme? identityref | +-- te-topology-identifier | | +-- provider-id? te-global-id | | +-- client-id? te-global-id | | +-- topology-id? te-topology-id | +-- te-bandwidth | | ... | +-- link-protection? identityref | +-- setup-priority? uint8 | +-- hold-priority? uint8 | +-- signaling-type? identityref | +-- hierarchy | +-- dependency-tunnels | | ... | +-- hierarchical-link | ...¶
The path requests that are intended to be used within the same tunnel should reference the same entry in the tunnel-attributes list. This allows:¶
- avoiding repeating the same set of per-tunnel parameters on multiple requested paths;¶
-
the server to understand what attributes are intended to be configured on a per-tunnel basis (e.g., the te-bandwidth configured in the tunnel-attributes) and what attributes are intended to be configured on a per-path basis(e.g., the te- bandwidth configured in the path-request). This could be useful especially when the server also creates a TE tunnel instance within the operational datastore to report the computed paths, as described in Section 3.3.1: in this case, the tunnel-name is also used as the suggested name for that TE tunnel instance.¶
The YANG data model allows also including requests for paths intended to modify existing tunnels (e.g., adding a protection path for an existing un-protected tunnel). In this case, the per-tunnel attributes are already provided in an existing TE tunnel instance and do not need to be re-configured in the path computation request RPC. Therefore, these requests should reference an existing TE tunnel instance.¶
It is also possible to request computing paths without indicating in which tunnel they are intended to be used (e.g., in case of an unprotected tunnel). In this case, the per-tunnel attributes could be provided together with the per-path attributes in the path request, without using the tunnel-attributes list.¶
The choices below are defined to distinguish the cases above:¶
-
whether the per-tunnel attributes are configured by reference (providing a leafref), to:¶
- or by value, providing the set of tunnel attributes within the path request:¶
| +-- (tunnel-attributes)? | | +--:(reference) | | | +-- tunnel-reference | | | +-- (tunnel-exist) | | | | +--:(tunnel-ref) | | | | | +-- tunnel-ref te:tunnel-ref | | | | +--:(tunnel-attributes-ref) | | | | +-- tunnel-attributes-ref leafref | | | +-- path-name? string | | | +-- (path-role) | | | ... | | +--:(value) | | +-- tunnel-name? string | | +-- path-name? string | | +-- (path-role)? | | | ... | | ... | | +-- encoding? identityref | | +-- switching-type? identityref | | ...¶
5.3.1. Tunnel attributes specified by value
The (value) case provides the set of attributes that are configured only on per-tunnel basis (e.g., tunnel-name, source/destination TTP, encoding and switching-type).¶
In this case, it is assumed that the requested path will be the only path within a tunnel.¶
If the only path within a tunnel is the transit segment of a multi-domain end-to-end path, it can be of any type (primary, secondary, reverse-primary or reverse-secondary). The (path-role) choice is used to specify its role in the path request:¶
| | +-- (path-role)? | | | +--:(primary-path) | | | +--:(secondary-path) | | | | +-- secondary-path! | | | | +-- primary-path-name? string | | | +--:(primary-reverse-path) | | | | +-- primary-reverse-path! | | | | +-- primary-path-name? string | | | +--:(secondary-reverse-path) | | | +-- secondary-reverse-path! | | | +-- primary-path-name? string | | | +-- primary-reverse-path-name? string¶
In all the other cases, the only path within a tunnel is a primary path.¶
The secondary-path, the primary-reverse-path and the secondary- reverse-path are presence container used to indicate the role of the path: by default, the path is assumed to be a primary path.¶
They optionally can contain the name of the primary-path under which the requested path could be associated within the YANG tree structure defined in [I-D.ietf-teas-yang-te], which could be useful when the server also creates a TE tunnel instance within the operational datastore to report the computed paths, as described in Section 3.3.1: in this case, the tunnel-name and the path names are also used as the suggested name for that TE tunnel and path instances.¶
5.3.2. Tunnel attributes specified by reference
The (reference) case provides the information needed to associate multiple path requests that are intended to be used within the same tunnel.¶
In order to indicate the role of the path being requested within the intended tunnel (e.g., primary or secondary path), the (path-role) choice is defined:¶
| | | +-- (path-role) | | | +--:(primary-path) | | | | +-- primary-path! | | | | ... | | | +--:(secondary-path) | | | | +-- secondary-path | | | | ... | | | +--:(primary-reverse-path) | | | | +-- primary-reverse-path | | | | ... | | | +--:(secondary-reverse-path) | | | +-- secondary-reverse-path | | | ...¶
The primary-path is a presence container used to indicate that the requested path is intended to be used as a primary path. It can also contain some attributes which are configured only on primary paths (e.g., the k-requested-paths).¶
The secondary-path container indicates that the requested path is intended to be used as a secondary path and it contains at least references to one or more primary paths which can use it as a candidate secondary path:¶
| | | | +-- secondary-path | | | | ... | | | | +-- primary-path-ref* [] | | | | +-- (primary-path-exist) | | | | +--:(path-ref) | | | | | +-- primary-path-ref leafref | | | | +--:(path-request-ref) | | | | +-- path-request-ref leafref¶
A requested secondary path can reference any requested primary paths, and, in case they are intended to be used within an existing TE tunnel, it could also reference any existing primary-paths.¶
The secondary-path container can also contain some attributes which are configured only on secondary paths (e.g., the protection-type).¶
The primary-reverse-path container indicates that the requested path is intended to be used as a primary reverse path and it contains only the reference to the primary path which is intended to use it as a reverse path:¶
| | | | +-- primary-reverse-path | | | | +-- (primary-path-exist) | | | | +--:(path-ref) | | | | | +-- primary-path-ref leafref | | | | +--:(path-request-ref) | | | | +-- path-request-ref leafref¶
A requested primary reverse path can reference either a requested primary path, or, in case it is intended to be used within an existing TE tunnel, an existing primary-path.¶
The secondary-reverse-path container indicates that the requested path is intended to be used as a secondary reverse path and it contains at least references to one or more primary paths, whose primary reverse path can use it as a candidate secondary reverse path:¶
| | | +-- secondary-reverse-path | | | ... | | | +-- primary-reverse-path* [] | | | +-- (primary-reverse-path-exist) | | | +--:(path-ref) | | | | +-- primary-path-ref leafref | | | +--:(path-request-ref) | | | +-- path-request-ref leafref¶
A requested secondary reverse path can reference any requested primary paths, and, in case they are intended to be used within an existing TE tunnel, it could reference also existing primary-paths.¶
The secondary-reverse-path container can also contain some attributes which are configured only on secondary reverse paths (e.g., the protection-type).¶
In case the requested path is a transit segment of a multi-domain end-to-end path, the primary-path may not be needed to be setup/computed. However, the request for path computation of a secondary-path or a primary-reverse or of a secondary-reverse-path requires that the primary-path exists or it is requested within the same RPC request. In the latter case, the path request for the primary-path should have an empty 'route-object-include-exclude' list, as described in section 5.1.1 of [I-D.ietf-teas-yang-te], to indicate to the server that path computation is not requested and no path properties will therefore be returned in the RPC response.¶
5.4. Multi-Layer Path Computation
The models supports requesting multi-layer path computation following the same approach based on dependency tunnels, as defined in [I-D.ietf-teas-yang-te].¶
The tunnel-attributes of a given client-layer path request can reference server-layer TE tunnels which can already exist in the YANG datastore or be specified in the tunnel-attributes list, within the same RPC request:¶
| +-- dependency-tunnels | | +-- dependency-tunnel* [name] | | | +-- name -> /te:te/tunnels/tunnel/name | | | +-- encoding? identityref | | | +-- switching-type? identityref | | +-- dependency-tunnel-attributes* [name] | | +-- name leafref | | +-- encoding? identityref | | +-- switching-type? identityref¶
In a similar way as in [I-D.ietf-teas-yang-te], the server-layer tunnel attributes should provide the information of what would be the dynamic link in the client layer topology supported by that tunnel, if instantiated:¶
| +-- hierarchical-link | +-- enable? boolean | +-- local-te-node-id? te-types:te-node-id | +-- local-te-link-tp-id? te-types:te-tp-id | +-- remote-te-node-id? te-types:te-node-id | +-- te-topology-identifier | +-- provider-id? te-global-id | +-- client-id? te-global-id | +-- topology-id? te-topology-id¶
It is worth noting that since path computation RPC is stateless, the dynamic hierarchical links configured for the server-layer tunnel attributes cannot be used for path computation of any client-layer path unless explicitly referenced in the dependency-tunnel-attributes list within the same RPC request.¶
6. YANG data model for TE path computation
6.1. Tree diagram
Figure 11 below shows the tree diagram of the YANG data model defined in module ietf-te-path-computation.yang, defined in Section 6.2.¶
7. Security Considerations
This document describes use cases of requesting Path Computation using YANG data models, which could be used at the ABNO Control Interface [RFC7491] and/or between controllers in ACTN [RFC8453]. As such, it does not introduce any new security considerations compared to the ones related to YANG specification, ABNO specification and ACTN Framework defined in [RFC7950], [RFC7491] and [RFC8453].¶
The YANG module defined in this document is designed to be accessed via the NETCONF protocol [RFC6241] or RESTCONF protocol [RFC8040]. The lowest NETCONF layer is the secure transport layer, and the mandatory-to-implement secure transport is Secure Shell (SSH) [RFC6242]. The lowest RESTCONF layer is HTTPS, and the mandatory-to- implement secure transport is TLS [RFC8446].¶
The Network Configuration Access Control Model (NACM) [RFC8341] provides the means to restrict access for particular NETCONF or RESTCONF users to a preconfigured subset of all available NETCONF or RESTCONF protocol operations and content.¶
The YANG module defined in this document augments the "tunnels-path-compute" and the "tunnel-actions" RPCs defined in [I-D.ietf-teas-yang-te]. The security considerations provided in [I-D.ietf-teas-yang-te] are also applicable to the YANG module defined in this document.¶
The RPC defined in this document can also be used for Denial-of-service (DoS) attacks. The security considerations defines in section 10.7.2 of [RFC5440] also applies to the use of this RPC.¶
The definition of the input shaping/policing mechanisms and of their configuration is outside the scope of this document.¶
Some of the RPC operations defined in this YANG module may be considered sensitive or vulnerable in some network environments. It is thus important to control access to these operations. These are the operations and their sensitivity/vulnerability:¶
"te-pc:response/computed-paths-properties": provides the same information provided by the "te:computed-paths-properties" defined in [I-D.ietf-teas-yang-te]. The security considerations provided in [I-D.ietf-teas-yang-te] for the TE tunnel state apply also to this subtree.¶
"te-pc:response/te-pc:tunnel-ref", "te-pc:response/te-pc:primary-path-ref", "te-pc:response/te-pc:primary-reverse-path-ref", "te-pc:response/te-pc:secondary-path-ref" and "te-pc:response/te-pc:secondary-reverse-path-ref" provides a reference where the same information provided in "te-pc:response/computed-paths-properties" is temporarly stored with the operational datastore (see Section 3.3.1). Therefore access to this information does not provide any additional security issue that the information provided with "te-pc:response/computed-paths-properties".¶
"/te:tunnels-actions": the YANG model defined in this document augments this action with a new action type that allows deleting the transient states of computed paths (see Section 3.3.1). A malicious use of this action would have no impact on the paths carrying live traffic but it would preclude the client from using the "transient states" to request the set-up of exactly that path, if still available.¶
The security considerations spelled out in the YANG specification [RFC7950] apply for this document as well.¶
8. IANA Considerations
This document registers the following URIs in the "ns" subregistry within the "IETF XML registry" [RFC3688].¶
URI: urn:ietf:params:xml:ns:yang:ietf-te-path-computation Registrant Contact: The IESG. XML: N/A, the requested URI is an XML namespace.¶
This document registers a YANG module in the "YANG Module Names" registry [RFC7950].¶
name: ietf-te-path-computation namespace: urn:ietf:params:xml:ns:yang:ietf-te-path-computation prefix: te-pc reference: this document¶
9. References
9.1. Normative References
- [I-D.ietf-teas-rfc8776-update]
- Busi, I., Guo, A., Liu, X., Saad, T., Beeram, V. P., and I. Bryskin, "Common YANG Data Types for Traffic Engineering", Work in Progress, Internet-Draft, draft-ietf-teas-rfc8776-update-04, , <https://datatracker.ietf.org/doc/html/draft-ietf-teas-rfc8776-update-04>.
- [I-D.ietf-teas-yang-te]
- Saad, T., Gandhi, R., Liu, X., Beeram, V. P., Bryskin, I., and O. G. de Dios, "A YANG Data Model for Traffic Engineering Tunnels, Label Switched Paths and Interfaces", Work in Progress, Internet-Draft, draft-ietf-teas-yang-te-33, , <https://datatracker.ietf.org/doc/html/draft-ietf-teas-yang-te-33>.
- [RFC2119]
- Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, DOI 10.17487/RFC2119, , <https://www.rfc-editor.org/info/rfc2119>.
- [RFC3688]
- Mealling, M., "The IETF XML Registry", BCP 81, RFC 3688, DOI 10.17487/RFC3688, , <https://www.rfc-editor.org/info/rfc3688>.
- [RFC5440]
- Vasseur, JP., Ed. and JL. Le Roux, Ed., "Path Computation Element (PCE) Communication Protocol (PCEP)", RFC 5440, DOI 10.17487/RFC5440, , <https://www.rfc-editor.org/info/rfc5440>.
- [RFC5441]
- Vasseur, JP., Ed., Zhang, R., Bitar, N., and JL. Le Roux, "A Backward-Recursive PCE-Based Computation (BRPC) Procedure to Compute Shortest Constrained Inter-Domain Traffic Engineering Label Switched Paths", RFC 5441, DOI 10.17487/RFC5441, , <https://www.rfc-editor.org/info/rfc5441>.
- [RFC5541]
- Le Roux, JL., Vasseur, JP., and Y. Lee, "Encoding of Objective Functions in the Path Computation Element Communication Protocol (PCEP)", RFC 5541, DOI 10.17487/RFC5541, , <https://www.rfc-editor.org/info/rfc5541>.
- [RFC6241]
- Enns, R., Ed., Bjorklund, M., Ed., Schoenwaelder, J., Ed., and A. Bierman, Ed., "Network Configuration Protocol (NETCONF)", RFC 6241, DOI 10.17487/RFC6241, , <https://www.rfc-editor.org/info/rfc6241>.
- [RFC6242]
- Wasserman, M., "Using the NETCONF Protocol over Secure Shell (SSH)", RFC 6242, DOI 10.17487/RFC6242, , <https://www.rfc-editor.org/info/rfc6242>.
- [RFC7926]
- Farrel, A., Ed., Drake, J., Bitar, N., Swallow, G., Ceccarelli, D., and X. Zhang, "Problem Statement and Architecture for Information Exchange between Interconnected Traffic-Engineered Networks", BCP 206, RFC 7926, DOI 10.17487/RFC7926, , <https://www.rfc-editor.org/info/rfc7926>.
- [RFC7950]
- Bjorklund, M., Ed., "The YANG 1.1 Data Modeling Language", RFC 7950, DOI 10.17487/RFC7950, , <https://www.rfc-editor.org/info/rfc7950>.
- [RFC8040]
- Bierman, A., Bjorklund, M., and K. Watsen, "RESTCONF Protocol", RFC 8040, DOI 10.17487/RFC8040, , <https://www.rfc-editor.org/info/rfc8040>.
- [RFC8174]
- Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC 2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174, , <https://www.rfc-editor.org/info/rfc8174>.
- [RFC8340]
- Bjorklund, M. and L. Berger, Ed., "YANG Tree Diagrams", BCP 215, RFC 8340, DOI 10.17487/RFC8340, , <https://www.rfc-editor.org/info/rfc8340>.
- [RFC8341]
- Bierman, A. and M. Bjorklund, "Network Configuration Access Control Model", STD 91, RFC 8341, DOI 10.17487/RFC8341, , <https://www.rfc-editor.org/info/rfc8341>.
- [RFC8446]
- Rescorla, E., "The Transport Layer Security (TLS) Protocol Version 1.3", RFC 8446, DOI 10.17487/RFC8446, , <https://www.rfc-editor.org/info/rfc8446>.
- [RFC8795]
- Liu, X., Bryskin, I., Beeram, V., Saad, T., Shah, H., and O. Gonzalez de Dios, "YANG Data Model for Traffic Engineering (TE) Topologies", RFC 8795, DOI 10.17487/RFC8795, , <https://www.rfc-editor.org/info/rfc8795>.
9.2. Informative References
- [I-D.ietf-ccamp-otn-topo-yang]
- Zheng, H., Busi, I., Liu, X., Belotti, S., and O. G. de Dios, "A YANG Data Model for Optical Transport Network Topology", Work in Progress, Internet-Draft, draft-ietf-ccamp-otn-topo-yang-16, , <https://datatracker.ietf.org/doc/html/draft-ietf-ccamp-otn-topo-yang-16>.
- [RFC4655]
- Farrel, A., Vasseur, J.-P., and J. Ash, "A Path Computation Element (PCE)-Based Architecture", RFC 4655, DOI 10.17487/RFC4655, , <https://www.rfc-editor.org/info/rfc4655>.
- [RFC6805]
- King, D., Ed. and A. Farrel, Ed., "The Application of the Path Computation Element Architecture to the Determination of a Sequence of Domains in MPLS and GMPLS", RFC 6805, DOI 10.17487/RFC6805, , <https://www.rfc-editor.org/info/rfc6805>.
- [RFC7446]
- Lee, Y., Ed., Bernstein, G., Ed., Li, D., and W. Imajuku, "Routing and Wavelength Assignment Information Model for Wavelength Switched Optical Networks", RFC 7446, DOI 10.17487/RFC7446, , <https://www.rfc-editor.org/info/rfc7446>.
- [RFC7491]
- King, D. and A. Farrel, "A PCE-Based Architecture for Application-Based Network Operations", RFC 7491, DOI 10.17487/RFC7491, , <https://www.rfc-editor.org/info/rfc7491>.
- [RFC8342]
- Bjorklund, M., Schoenwaelder, J., Shafer, P., Watsen, K., and R. Wilton, "Network Management Datastore Architecture (NMDA)", RFC 8342, DOI 10.17487/RFC8342, , <https://www.rfc-editor.org/info/rfc8342>.
- [RFC8453]
- Ceccarelli, D., Ed. and Y. Lee, Ed., "Framework for Abstraction and Control of TE Networks (ACTN)", RFC 8453, DOI 10.17487/RFC8453, , <https://www.rfc-editor.org/info/rfc8453>.
- [RFC8454]
- Lee, Y., Belotti, S., Dhody, D., Ceccarelli, D., and B. Yoon, "Information Model for Abstraction and Control of TE Networks (ACTN)", RFC 8454, DOI 10.17487/RFC8454, , <https://www.rfc-editor.org/info/rfc8454>.
Appendix A. Examples
This section contains examples of use of the model with RESTCONF [RFC8040] and JSON encoding.¶
These examples show how path computation can be requested for the tunnels configuration provided in Appendix A of [I-D.ietf-teas-yang-te].¶
A.1. Basic Path Computation
This example uses the path computation RPC defined in this document to request the computation of the path for the tunnel defined in section 12.1 of of [I-D.ietf-teas-yang-te].¶
In this case, the TE Tunnel has only one primary path with no specific constraints.¶
POST /restconf/operations/ietf-te:tunnels-path-compute HTTP/1.1 Host: example.com Content-Type: application/yang-data+json¶
{ "ietf-te:input": { "path-compute-info": { "ietf-te-path-computation:path-request": [ { "request-id": 1, "tunnel-name": "Example_LSP_Tunnel_A_2", "encoding": "te-types:lsp-encoding-packet", "source": "192.0.2.1", "destination": "192.0.2.4", "bidirectional": "false", "signaling-type": "te-types:path-setup-rsvp" } ] } } }¶
A.2. Path Computation with transient state
This example uses the path computation RPC defined in this document to request the computation of the path for the tunnel defined in section 12.1 of of [I-D.ietf-teas-yang-te] requesting some transient state to be reported within the operational datastore, as described Section 3.3.1.¶
In this case, the TE Tunnel has only one primary path with no specific constraints.¶
POST /restconf/operations/ietf-te:tunnels-path-compute HTTP/1.1 Host: example.com Content-Type: application/yang-data+json¶
{ "ietf-te:input": { "path-compute-info": { "ietf-te-path-computation:path-request": [ { "request-id": 2, "tunnel-name": "Example_LSP_Tunnel_A_2", "encoding": "te-types:lsp-encoding-packet", "source": "192.0.2.1", "destination": "192.0.2.4", "bidirectional": "false", "signaling-type": "te-types:path-setup-rsvp", "requested-state": { "transaction-id": "example" } } ] } } }¶
A.3. Path Computation with Global Path Constraint
This example uses the path computation RPC defined in this document to request the computation of the path for the tunnel defined in section 12.3 of of [I-D.ietf-teas-yang-te]. The 'named path constraint' is created in section 12.2 of [I-D.ietf-teas-yang-te] applies to this path computation request.¶
POST /restconf/operations/ietf-te:tunnels-path-compute HTTP/1.1 Host: example.com Content-Type: application/yang-data+json¶
{ "ietf-te:input": { "path-compute-info": { "ietf-te-path-computation:path-request": [ { "request-id": 3, "tunnel-name": "Example_LSP_Tunnel_A_4_1", "path-name": "Simple_LSP_1", "encoding": "te-types:lsp-encoding-packet", "source": "192.0.2.1", "destination": "192.0.2.4", "bidirectional": "false", "signaling-type": "path-setup-rsvp", "named-path-constraint": "max-hop-3", "requested-state": {} } ] } } }¶
A.4. Path Computation with Per-tunnel Path Constraint
This example uses the path computation RPC defined in this document to request the computation of the path for the tunnel defined in section 12.4 of of [I-D.ietf-teas-yang-te], using a per tunnel path constraint.¶
POST /restconf/operations/ietf-te:tunnels-path-compute HTTP/1.1 Host: example.com Content-Type: application/yang-data+json¶
{ "ietf-te:input": { "path-compute-info": { "ietf-te-path-computation:path-request": [ { "request-id": 4, "tunnel-name": "Example_LSP_Tunnel_A_4_2", "path-name": "path1", "encoding": "te-types:lsp-encoding-packet", "source": "192.0.2.1", "destination": "192.0.2.4", "bidirectional": "false", "signaling-type": "te-types:path-setup-rsvp", "path-metric-bounds": { "path-metric-bound": [ { "metric-type": "te-types:path-metric-hop", "upper-bound": "3" } ] } } ] } } }¶
A.5. Path Computation result
This example reports the output of the path computation RPC request described in Appendix A.4.¶
HTTP/1.1 200 OK Host: example.com Content-Type: application/yang-data+json¶
{ "ietf-te:output": { "path-compute-result": { "ietf-te-path-computation:response": [ { "response-id": 3, "computed-paths-properties": { "computed-path-properties": [ { "k-index": "1", "path-properties": { "path-route-objects": { "path-route-object": [ { "index": "1", "numbered-node-hop": { "node-id": "192.0.2.2" } }, { "index": "2", "numbered-node-hop": { "node-id": "192.0.2.4" } } ] } } } ] }, "tunnel-ref": "Example_LSP_Tunnel_A_4_1", "primary-path-ref": "path1" } ] } } }¶
A.6. Path Computation with Primary and Secondary Paths
This section contains examples of use of the model to compute primary and secondary paths described in section 12.6 of [I-D.ietf-teas-yang-te].¶
These examples consider the cases where: - primary and reverse paths are unidirectional and not co-routed (example-1); - primary and reverse paths are bidirectional (example-3); - primary and reverse paths are unidirectional and co-routed (example-4).¶
POST /restconf/operations/ietf-te:tunnels-path-compute HTTP/1.1 Host: example.com Content-Type: application/yang-data+json¶
{ "ietf-te:input": { "path-compute-info": { "ietf-te-path-computation:path-request": [ { "request-id": 1, "tunnel-reference": { "tunnel-attributes-ref": "example-1", "path-name": "primary-1 (fwd)", "primary-path": {} }, "explicit-route-objects-always": { "route-object-include-exclude": [ { "index": 1, "numbered-node-hop": { "node-id": "192.0.2.2", "hop-type": "loose" } } ] } }, { "request-id": 2, "tunnel-reference": { "tunnel-attributes-ref": "example-1", "path-name": "primary-2 (rev)", "primary-reverse-path": { "primary-path-request-ref": 1 } }, "explicit-route-objects-always": { "route-object-include-exclude": [ { "index": 1, "numbered-node-hop": { "node-id": "192.0.2.3", "hop-type": "loose" } } ] } }, { "request-id": 3, "tunnel-reference": { "tunnel-attributes-ref": "example-1", "path-name": "secondary-1 (fwd)", "secondary-path": { "primary-path": [ { "path-request-ref": 1 } ] } }, "explicit-route-objects-always": { "route-object-include-exclude": [ { "index": 1, "numbered-node-hop": { "node-id": "192.0.2.1" } }, { "index": 2, "numbered-node-hop": { "node-id": "192.0.2.2", "hop-type": "loose" } } ] } }, { "request-id": 4, "tunnel-reference": { "tunnel-attributes-ref": "example-1", "path-name": "secondary-2 (fwd)", "secondary-path": { "primary-path": [ { "path-request-ref": 1 } ] } }, "explicit-route-objects-always": { "route-object-include-exclude": [ { "index": 1, "numbered-node-hop": { "node-id": "192.0.2.2" } }, { "index": 2, "numbered-node-hop": { "node-id": "192.0.2.5", "hop-type": "loose" } } ] } }, { "request-id": 5, "tunnel-reference": { "tunnel-attributes-ref": "example-1", "path-name": "secondary-3 (rev)", "secondary-reverse-path": { "primary-reverse-path": [ { "path-request-ref": 2 } ] } }, "explicit-route-objects-always": { "route-object-include-exclude": [ { "index": 1, "numbered-node-hop": { "node-id": "192.0.2.5" } }, { "index": 2, "numbered-node-hop": { "node-id": "192.0.2.4", "hop-type": "loose" } } ] } }, { "request-id": 6, "tunnel-reference": { "tunnel-attributes-ref": "example-1", "path-name": "secondary-4 (rev)", "secondary-reverse-path": { "primary-reverse-path": [ { "path-request-ref": 2 } ] } }, "explicit-route-objects-always": { "route-object-include-exclude": [ { "index": 1, "numbered-node-hop": { "node-id": "192.0.2.4" } }, { "index": 2, "numbered-node-hop": { "node-id": "192.0.2.3", "hop-type": "loose" } } ] } }, { "request-id": 7, "tunnel-reference": { "tunnel-attributes-ref": "example-1", "path-name": "secondary-5 (rev)", "secondary-reverse-path": { "primary-reverse-path": [ { "path-request-ref": 2 } ] } }, "explicit-route-objects-always": { "route-object-include-exclude": [ { "index": 1, "numbered-node-hop": { "node-id": "192.0.2.3" } }, { "index": 2, "numbered-node-hop": { "node-id": "192.0.2.1", "hop-type": "loose" } } ] } }, { "request-id": 8, "tunnel-reference": { "tunnel-attributes-ref": "example-3", "path-name": "primary-1 (bidir)", "primary-path": {} }, "explicit-route-objects-always": { "route-object-include-exclude": [ { "index": 1, "numbered-node-hop": { "node-id": "192.0.2.2", "hop-type": "loose" } } ] } }, { "request-id": 9, "tunnel-reference": { "tunnel-attributes-ref": "example-3", "path-name": "secondary-1 (bidir)", "secondary-path": { "primary-path": [ { "path-request-ref": 8 } ] } }, "explicit-route-objects-always": { "route-object-include-exclude": [ { "index": 1, "numbered-node-hop": { "node-id": "192.0.2.1" } }, { "index": 2, "numbered-node-hop": { "node-id": "192.0.2.2", "hop-type": "loose" } } ] } }, { "request-id": 10, "tunnel-reference": { "tunnel-attributes-ref": "example-3", "path-name": "secondary-2 (bidir)", "secondary-path": { "primary-path": [ { "path-request-ref": 8 } ] } }, "explicit-route-objects-always": { "route-object-include-exclude": [ { "index": 1, "numbered-node-hop": { "node-id": "192.0.2.2" } }, { "index": 2, "numbered-node-hop": { "node-id": "192.0.2.5", "hop-type": "loose" } } ] } }, { "request-id": 11, "tunnel-reference": { "tunnel-attributes-ref": "example-4", "path-name": "primary-1 (fwd)", "primary-path": { "co-routed": [null] } }, "explicit-route-objects-always": { "route-object-include-exclude": [ { "index": 1, "numbered-node-hop": { "node-id": "192.0.2.2", "hop-type": "loose" } } ] } }, { "request-id": 12, "tunnel-reference": { "tunnel-attributes-ref": "example-4", "path-name": "primary-2 (rev)", "primary-reverse-path": { "primary-path-request-ref": 11 } } }, { "request-id": 13, "tunnel-reference": { "tunnel-attributes-ref": "example-4", "path-name": "secondary-1 (fwd)", "secondary-path": { "primary-path": [ { "co-routed": [null], "path-request-ref": 11 } ] } }, "explicit-route-objects-always": { "route-object-include-exclude": [ { "index": 1, "numbered-node-hop": { "node-id": "192.0.2.1" } }, { "index": 2, "numbered-node-hop": { "node-id": "192.0.2.2", "hop-type": "loose" } } ] } }, { "request-id": 14, "tunnel-reference": { "tunnel-attributes-ref": "example-4", "path-name": "secondary-2 (fwd)", "secondary-path": { "primary-path": [ { "co-routed": [null], "path-request-ref": 11 } ] } }, "explicit-route-objects-always": { "route-object-include-exclude": [ { "index": 1, "numbered-node-hop": { "node-id": "192.0.2.2" } }, { "index": 2, "numbered-node-hop": { "node-id": "192.0.2.5", "hop-type": "loose" } } ] } }, { "request-id": 15, "tunnel-reference": { "tunnel-attributes-ref": "example-1", "path-name": "secondary-3 (rev)", "secondary-reverse-path": { "primary-reverse-path": [ { "path-request-ref": 12 } ] } } }, { "request-id": 16, "tunnel-reference": { "tunnel-attributes-ref": "example-1", "path-name": "secondary-4 (rev)", "secondary-reverse-path": { "primary-reverse-path": [ { "path-request-ref": 12 } ] } } } ], "ietf-te-path-computation:tunnel-attributes": [ { "tunnel-name": "example-1", "source": "192.0.2.1", "destination": "192.0.2.5", "bidirectional": "false" }, { "tunnel-name": "example-3", "source": "192.0.2.1", "destination": "192.0.2.5", "bidirectional": "true" }, { "tunnel-name": "example-4", "source": "192.0.2.1", "destination": "192.0.2.5", "bidirectional": "false" } ] } } }¶
Acknowledgments
The authors would like to thank Igor Bryskin and Xian Zhang for participating in the initial discussions that have triggered this work and providing valuable insights.¶
The authors would like to thank the authors of the TE tunnel YANG data model [I-D.ietf-teas-yang-te], in particular Igor Bryskin, Vishnu Pavan Beeram, Tarek Saad and Xufeng Liu, for their inputs to the discussions and support in having consistency between the Path Computation and TE tunnel YANG data models.¶
The authors would like to thank Adrian Farrel, Dhruv Dhody, Igor Bryskin, Julien Meuric and Lou Berger for their valuable input to the discussions that has clarified that the path being set up is not necessarily the same as the path that has been previously computed and, in particular to Dhruv Dhody, for his suggestion to describe the need for a path verification phase to check that the actual path being set up meets the required end-to-end metrics and constraints.¶
The authors would like to thank Aihua Guo, Lou Berger, Shaolong Gan, Martin Bjorklund and Tom Petch for their useful comments on how to define XPath statements in YANG RPCs.¶
The authors would like to thank Haomian Zheng, Yanlei Zheng, Tom Petch, Aihua Guo and Martin Bjorklund for their review and valuable comments to this document.¶
Previous versions of document were prepared using 2-Word-v2.0.template.dot.¶
This document was prepared using kramdown.¶