| RFC 8751 | Hierarchical Stateful PCE | March 2020 |
| Dhody, et al. | Informational | [Page] |
A stateful Path Computation Element (PCE) maintains information on the current network state received from the Path Computation Clients (PCCs), including computed Label Switched Paths (LSPs), reserved resources within the network, and pending path computation requests. This information may then be considered when computing the path for a new traffic-engineered LSP or for any associated/dependent LSPs. The path-computation response from a PCE helps the PCC to gracefully establish the computed LSP.¶
The Hierarchical Path Computation Element (H-PCE) architecture allows the optimum sequence of interconnected domains to be selected and network policy to be applied if applicable, via the use of a hierarchical relationship between PCEs.¶
Combining the capabilities of stateful PCE and the hierarchical PCE would be advantageous. This document describes general considerations and use cases for the deployment of stateful, but not stateless, PCEs using the hierarchical PCE architecture.¶
This document is not an Internet Standards Track specification; it is published for informational purposes.¶
This document is a product of the Internet Engineering Task Force (IETF). It represents the consensus of the IETF community. It has received public review and has been approved for publication by the Internet Engineering Steering Group (IESG). Not all documents approved by the IESG are candidates for any level of Internet Standard; see Section 2 of RFC 7841.¶
Information about the current status of this document, any errata, and how to provide feedback on it may be obtained at https://www.rfc-editor.org/info/rfc8751.¶
Copyright (c) 2020 IETF Trust and the persons identified as the document authors. All rights reserved.¶
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License.¶
The Path Computation Element communication Protocol (PCEP) [RFC5440] provides mechanisms for Path Computation Elements (PCEs) to perform path computations in response to the requests of Path Computation Clients (PCCs).¶
A stateful PCE is capable of considering, for the purposes of path computation, not only the network state in terms of links and nodes (referred to as the Traffic Engineering Database or TED) but also the status of active services (previously computed paths, and currently reserved resources, stored in the Label Switched Paths Database (LSPDB).¶
[RFC8051] describes general considerations for a stateful PCE deployment; it also examines its applicability and benefits as well as its challenges and limitations through a number of use cases.¶
[RFC8231] describes a set of extensions to PCEP to provide stateful control. For its computations, a stateful PCE has access to not only the information carried by the network's Interior Gateway Protocol (IGP), but also the set of active paths and their reserved resources. The additional state allows the PCE to compute constrained paths while considering individual LSPs and their interactions. [RFC8281] describes the setup, maintenance, and teardown of PCE-initiated LSPs under the stateful PCE model.¶
[RFC8231] also describes the active stateful PCE. The active PCE functionality allows a PCE to reroute an existing LSP, make changes to the attributes of an existing LSP, or delegate control of specific LSPs to a new PCE.¶
The ability to compute constrained paths for Traffic Engineering (TE) LSPs in Multiprotocol Label Switching (MPLS) and Generalized MPLS (GMPLS) networks across multiple domains has been identified as a key motivation for PCE development. [RFC6805] describes a Hierarchical PCE (H-PCE) architecture that can be used for computing end-to-end paths for interdomain MPLS TE and GMPLS Label Switched Paths (LSPs). Within the H-PCE architecture [RFC6805], the Parent PCE (P-PCE) is used to compute a multidomain path based on the domain connectivity information. A Child PCE (C-PCE) may be responsible for a single domain or multiple domains. The C-PCE is used to compute the intradomain path based on its domain topology information.¶
This document presents general considerations for stateful PCEs, and not stateless PCEs, in the hierarchical PCE architecture. It focuses on the behavior changes and additions to the existing stateful PCE mechanisms (including PCE-initiated LSP setup and active stateful PCE usage) in the context of networks using the H-PCE architecture.¶
In this document, Sections 3.1 and 3.2 focus on end-to-end (E2E) interdomain TE LSP. Section 3.3.1 describes the operations for stitching per-domain LSPs.¶
As per [RFC6805], in the hierarchical PCE architecture, a P-PCE maintains a domain topology map that contains the child domains and their interconnections. Usually, the P-PCE has no information about the content of the child domains. But, if the PCE is applied to the Abstraction and Control of TE Networks (ACTN) [RFC8453] as described in [RFC8637], the Provisioning Network Controller (PNC) can provide an abstract topology to the Multi-Domain Service Coordinator (MDSC). Thus, the P-PCE in MDSC could be aware of topology information in much more detail than just the domain topology.¶
In a PCEP session between a PCC (ingress) and a C-PCE, the C-PCE acts as per the stateful PCE operations described in [RFC8231] and [RFC8281]. The same C-PCE behaves as a PCC on the PCEP session towards the P-PCE. The P-PCE is stateful in nature; thus, it maintains the state of the interdomain LSPs that are reported to it. The interdomain LSP could also be delegated by the C-PCE to the P-PCE, so that the P-PCE could update the interdomain path. The trigger for this update could be the LSP state change reported for this LSP or any other LSP. It could also be a change in topology at the P-PCE, such as interdomain link status change. In case of use of stateful H-PCE in ACTN, a change in abstract topology learned by the P-PCE could also trigger the update. Some other external factors (such as a measurement probe) could also be a trigger at the P-PCE. Any such update would require an interdomain path recomputation as described in [RFC6805].¶
The end-to-end interdomain path computation and setup is described in [RFC6805]. Additionally, a per-domain stitched-LSP model is also applicable in a P-PCE initiation model. Sections 3.1, 3.2, and 3.3 describe the end-to-end contiguous LSP setup, whereas Section 3.3.1 describes the per-domain stitching.¶
[RFC8453] describes a framework for the Abstraction and Control of TE Networks (ACTN), where each Provisioning Network Controller (PNC) is equivalent to a C-PCE, and the P-PCE is the Multi-Domain Service Coordinator (MDSC). The per-domain stitched LSP is well suited for ACTN deployments, as per the hierarchical PCE architecture described in Section 3.3.1 of this document and Section 4.1 of [RFC8453].¶
[RFC8637] examines the applicability of PCE to the ACTN framework. To support the function of multidomain coordination via hierarchy, the hierarchy of stateful PCEs plays a crucial role.¶
In the ACTN framework, a Customer Network Controller (CNC) can request the MDSC to check whether there is a possibility to meet Virtual Network (VN) requirements before requesting that the VN be provisioned. The H-PCE architecture as described in [RFC6805] can support this function using Path Computation Request and Reply (PCReq and PCRep, respectively) messages between the P-PCE and C-PCEs. When the CNC requests VN provisioning, the MDSC decomposes this request into multiple interdomain LSP provisioning requests, which might be further decomposed into per-domain path segments. This is described in Section 3.3.1. The MDSC uses the LSP initiate request (PCInitiate) message from the P-PCE towards the C-PCE, and the C-PCE reports the state back to the P-PCE via a Path Computation State Report (PCRpt) message. The P-PCE could make changes to the LSP via the use of a Path Computation Update Request (PCUpd) message.¶
In this case, the P-PCE (as MDSC) interacts with multiple C-PCEs (as PNCs) along the interdomain path of the LSP.¶
Different signaling options for interdomain RSVP-TE are identified in [RFC4726]. Contiguous LSPs are achieved using the procedures of [RFC3209] and [RFC3473] to create a single end-to-end LSP that spans all domains. [RFC6805] describes the technique for establishing the optimum path when the sequence of domains is not known in advance.¶
That document shows how the PCE architecture can be extended to allow the optimum sequence of domains to be selected and the optimum end-to-end path to be derived.¶
A stateful P-PCE has to be aware of the interdomain LSPs for it to consider them during path computation. For instance, when a domain-diverse path is required from another LSP, the P-PCE needs to be aware of the LSP. This is the passive stateful P-PCE, as described in Section 3.1. Additionally, the interdomain LSP could be delegated to the P-PCE, so that P-PCE could trigger an update via a PCUpd message. The update could be triggered on receipt of the PCRpt message that indicates a status change of this LSP or some other LSP. The other LSP could be an associated LSP (such as a protection LSP [RFC8745]) or an unrelated LSP whose resource change leads to reoptimization at the P-PCE. This is the active stateful operation, as described in Section 3.2. Further, the P-PCE could be instructed to create an interdomain LSP on its own using the PCInitiate message for an E2E contiguous LSP. The P-PCE would send the PCInitiate message to the ingress domain C-PCE, which would further instruct the ingress PCC.¶
In this document, for the contiguous LSP, the above interactions are only between the ingress domain C-PCE and the P-PCE. The use of stateful operations for an interdomain LSP between the transit/egress domain C-PCEs and the P-PCE is out of the scope of this document.¶
[RFC8051] describes general considerations for a stateful PCE deployment and examines its applicability and benefits, as well as its challenges and limitations, through a number of use cases. These are also applicable to the stateful P-PCE when used for the interdomain LSP path computation and setup. It should be noted that though the stateful P-PCE has limited direct visibility inside the child domain, it could still trigger reoptimization with the help of child PCEs based on LSP state changes, abstract topology changes, or some other external factors.¶
The C-PCE would delegate control of the interdomain LSP to the P-PCE so that the P-PCE can make changes to it. Note that, if the C-PCE becomes aware of a topology change that is hidden from the P-PCE, it could take back the delegation from the P-PCE to act on it itself. Similarly, a P-PCE could also request delegation if it needs to make a change to the LSP (refer to [RFC8741]).¶
The terminology is as per [RFC4655], [RFC5440], [RFC6805], [RFC8051], [RFC8231], and [RFC8281].¶
Some key terms are listed below for easy reference.¶
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.¶
As described in [RFC6805], in the hierarchical PCE architecture, a P-PCE maintains a domain topology map that contains the child domains (seen as vertices in the topology) and their interconnections (links in the topology). Usually, the P-PCE has no information about the content of the child domains. Each child domain has at least one PCE capable of computing paths across the domain. These PCEs are known as Child PCEs (C-PCEs) [RFC6805] and have a direct relationship with the P-PCE. The P-PCE builds the domain topology map either via direct configuration or from learned information received from each C-PCE. The network policy could be applied while building the domain topology map. This has been described in detail in [RFC6805].¶
Note that, in the scope of this document, both the C-PCEs and the P-PCE are stateful in nature.¶
[RFC8231] specifies new functions to support a stateful PCE. It also specifies that a function can be initiated either from a PCC towards a PCE (C-E) or from a PCE towards a PCC (E-C).¶
This document extends these functions to support H-PCE Architecture from a C-PCE towards P-PCE (EC-EP) or from a P-PCE towards C-PCE (EP-EC). All PCE types herein (EC-EP and EP-EC) are assumed to be "stateful PCE".¶
A number of interactions are expected in the hierarchical stateful PCE architecture. These include:¶
Note that this hierarchy is recursive, so a Label Switching Router (LSR), as a PCC, could delegate control to a PCE. That PCE may, in turn, delegate to its parent, which may further delegate to its parent (if it exists). Similarly, update operations can also be applied recursively.¶
[RFC8685] defines the H-PCE-CAPABILITY TLV that is used in the Open message to advertise the H-PCE capability. [RFC8231] defines the STATEFUL-PCE-CAPABILITY TLV used in the Open message to indicate stateful support. To indicate the support for stateful H-PCE operations described in this document, a PCEP speaker MUST include both TLVs in an Open message. It is RECOMMENDED that any implementation that supports stateful operations [RFC8231] and H-PCE [RFC8685] also implement the stateful H-PCE operations as described in this document.¶
Further consideration may be made for optional procedures for stateful communication coordination between PCEs, including procedures to minimize computational loops. The procedures described in [PCE-STATE-SYNC] facilitate stateful communication between PCEs for various use cases. The procedures and extensions as described in Section 3 of [PCE-STATE-SYNC] are also applicable to child and parent PCE communication. The SPEAKER-IDENTITY-ID TLV (defined in [RFC8232]) is included in the LSP object to identify the ingress (PCC). The PCEP-specific identifier for the LSP (PLSP-ID [RFC8231]) used in the forwarded PCRpt by the C-PCE to the P-PCE is the same as the original one used by the PCC.¶
Procedures described in [RFC6805] are applied, where the ingress PCC triggers a path computation request for the destination towards the C-PCE in the domain where the LSP originates. The C-PCE further forwards the request to the P-PCE. The P-PCE selects a set of candidate domain paths based on the domain topology and the state of the interdomain links. It then sends computation requests to the C-PCEs responsible for each of the domains on the candidate domain paths. Each C-PCE computes a set of candidate path segments across its domain and sends the results to the P-PCE. The P-PCE uses this information to select path segments and concatenate them to derive the optimal end-to-end interdomain path. The end-to-end path is then sent to the C-PCE that received the initial path request, and this C-PCE passes the path on to the PCC that issued the original request.¶
As per [RFC8231], the PCC sends an LSP State Report carried on a PCRpt message to the C-PCE, indicating the LSP's status. The C-PCE may further propagate the State Report to the P-PCE. A local policy at the C-PCE may dictate which LSPs are reported to the P-PCE. The PCRpt message is sent from C-PCE to P-PCE.¶
State synchronization mechanisms as described in [RFC8231] and [RFC8232] are applicable to a PCEP session between C-PCE and P-PCE as well.¶
We use the hierarchical domain topology example from [RFC6805] as the reference topology for the entirety of this document. It is shown in Figure 1.¶
----------------------------------------------------------------- | Domain 5 | | ----- | | |PCE 5| | | ----- | | | | ---------------- ---------------- ---------------- | | | Domain 1 | | Domain 2 | | Domain 3 | | | | | | | | | | | | ----- | | ----- | | ----- | | | | |PCE 1| | | |PCE 2| | | |PCE 3| | | | | ----- | | ----- | | ----- | | | | | | | | | | | | ----| |---- ----| |---- | | | | |BN11+---+BN21| |BN23+---+BN31| | | | | - ----| |---- ----| |---- - | | | | |S| | | | | |D| | | | | - ----| |---- ----| |---- - | | | | |BN12+---+BN22| |BN24+---+BN32| | | | | ----| |---- ----| |---- | | | | | | | | | | | | ---- | | | | ---- | | | | |BN13| | | | | |BN33| | | | -----------+---- ---------------- ----+----------- | | \ / | | \ ---------------- / | | \ | | / | | \ |---- ----| / | | ----+BN41| |BN42+---- | | |---- ----| | | | | | | | ----- | | | | |PCE 4| | | | | ----- | | | | | | | | Domain 4 | | | ---------------- | | | -----------------------------------------------------------------
Steps 1 to 11 are exactly as described in Section 4.6.2 of [RFC6805] ("Hierarchical PCE End-to-End Path Computation Procedure"); the following additional steps are added for stateful PCE, to be executed at the end:¶
The ingress LSR could trigger path reoptimization by sending the path computation request as described in [RFC6805]; at this time, it can include the LSP object in the PCReq message, as described in [RFC8231].¶
[RFC8231] describes the case of an active stateful PCE. The active PCE functionality uses two specific PCEP messages:¶
The first is sent by the PCE to a PCC for modifying LSP attributes. The PCC sends back a PCRpt to acknowledge the requested operation or report any change in the LSP's state.¶
As per [RFC8051], delegation is an operation to grant a PCE temporary rights to modify a subset of LSP parameters on the LSPs of one or more PCCs. The C-PCE may further choose to delegate to its P-PCE based on a local policy. The PCRpt message with the "D" (delegate) flag is sent from C-PCE to P-PCE.¶
To update an LSP, a PCE sends an LSP Update Request to the PCC using a PCUpd message. For an LSP delegated to a P-PCE via the C-PCE, the P-PCE can use the same PCUpd message to request a change to the C-PCE (the ingress domain PCE). The C-PCE further propagates the update request to the PCC.¶
The P-PCE uses the same mechanism described in Section 3.1 to compute the end-to-end path using PCReq and PCRep messages.¶
For active operations, the following steps are required when delegating the LSP, again using the reference architecture described in Figure 1 ("Hierarchical Domain Topology Example").¶
[RFC8281] describes the setup, maintenance, and teardown of PCE-initiated LSPs under the stateful PCE model, without the need for local configuration on the PCC, thus allowing for a dynamic network that is centrally controlled and deployed. To instantiate or delete an LSP, the PCE sends the Path Computation LSP initiate request (PCInitiate) message to the PCC. In the case of an interdomain LSP in hierarchical PCE architecture, the initiation operations can be carried out at the P-PCE. In that case, after the P-PCE finishes the E2E path computation, it can send the PCInitiate message to the C-PCE (the ingress domain PCE), and the C-PCE further propagates the initiate request to the PCC.¶
The following steps are performed for PCE-initiated operations, again using the reference architecture described in Figure 1 ("Hierarchical Domain Topology Example"):¶
The ingress LSR (PCC) generates the PLSP-ID for the LSP and inform the C-PCE, which is propagated to the P-PCE.¶
The hierarchical PCE architecture, as per [RFC6805], is primarily used for E2E LSP. With PCE-initiated capability, another mode of operation is possible, where multiple intradomain LSPs are initiated in each domain and are further stitched to form an E2E LSP. The P-PCE sends PCInitiate message to each C-PCE separately to initiate individual LSP segments along the domain path. These individual per-domain LSPs are stitched together by some mechanism, which is out of the scope of this document (Refer to [STATEFUL-INTERDOMAIN]).¶
The following steps are performed for the per-domain stitched LSP operation, again using the reference architecture described in Figure 1 ("Hierarchical Domain Topology Example"):¶
The P-PCE (PCE5) is requested to initiate an LSP. Steps 4 to 10 in Section 4.6.2 of [RFC6805] are executed to determine the end-to-end path, which is broken into per-domain LSPs. For example:¶
It should be noted that the P-PCE may use other mechanisms to determine the suitable per-domain LSPs (apart from [RFC6805]).¶
For LSP (BN33-D):¶
For LSP (BN41-BN33):¶
For LSP (S-BN41):¶
Additionally:¶
Note that each per-domain LSP can be set up in parallel. Further, it is also possible to stitch the per-domain LSP at the same time as the per-domain LSPs are initiated. This option is defined in [STATEFUL-INTERDOMAIN].¶
The security considerations listed in [RFC8231], [RFC6805], and [RFC5440] apply to this document, as well. As per [RFC6805], it is expected that the parent PCE will require all child PCEs to use full security (i.e., the highest security mechanism available for PCEP) when communicating with the parent.¶
Any multidomain operation necessarily involves the exchange of information across domain boundaries. This is bound to represent a significant security and confidentiality risk, especially when the child domains are controlled by different commercial concerns. PCEP allows individual PCEs to maintain the confidentiality of their domain-path information using path-keys [RFC5520], and the hierarchical PCE architecture is specifically designed to enable as much isolation of information about domain topology and capabilities as is possible. The LSP state in the PCRpt message must continue to maintain the internal domain confidentiality when required.¶
The security considerations for PCE-initiated LSP in [RFC8281] are also applicable from P-PCE to C-PCE.¶
Further, Section 6.3 describes the use of a path-key [RFC5520] for confidentiality between C-PCE and P-PCE.¶
Thus, it is RECOMMENDED to secure the PCEP session (between the P-PCE and the C-PCE) using Transport Layer Security (TLS) [RFC8446] (per the recommendations and best current practices in BCP 195 [RFC7525]) and/or TCP Authentication Option (TCP-AO) [RFC5925]. The guidance for implementing PCEP with TLS can be found in [RFC8253].¶
In the case of TLS, due care needs to be taken while exposing the parameters of the X.509 certificate -- such as subjectAltName:otherName, which is set to Speaker Entity Identifier [RFC8232] as per [RFC8253] -- to ensure uniqueness and avoid any mismatch.¶
All manageability requirements and considerations listed in [RFC5440], [RFC6805], [RFC8231], and [RFC8281] apply to stateful H-PCE defined in this document. In addition, requirements and considerations listed in this section apply.¶
Support of the hierarchical procedure will be controlled by the management organization responsible for each child PCE. The parent PCE must only accept path-computation requests from authorized child PCEs. If a parent PCE receives a report from an unauthorized child PCE, the report should be dropped. All mechanisms described in [RFC8231] and [RFC8281] continue to apply.¶
An implementation should allow the operator to view the stateful and H-PCE capabilities advertised by each peer. The "ietf-pcep" PCEP YANG module is specified in [PCE-PCEP-YANG]. This YANG module will be required to be augmented to also include details for stateful H-PCE deployment and operation. The exact model and attributes are out of scope for this document.¶
Mechanisms defined in this document do not imply any new liveness-detection or monitoring requirements in addition to those already listed in [RFC5440].¶
Mechanisms defined in this document do not imply any new operation-verification requirements in addition to those already listed in [RFC5440] and [RFC8231].¶
Mechanisms defined in this document do not imply any new requirements on other protocols.¶
Mechanisms defined in [RFC5440] and [RFC8231] also apply to PCEP extensions defined in this document.¶
The stateful H-PCE technique brings the applicability of stateful PCE (described in [RFC8051]) to the LSP traversing multiple domains.¶
As described in Section 3, a PCEP speaker includes both the H-PCE-CAPABILITY TLV [RFC8685] and STATEFUL-PCE-CAPABILITY TLV [RFC8231] to indicate support for stateful H-PCE. Note that there is a possibility of a PCEP speaker that does not support the stateful H-PCE feature but does provide support for stateful-PCE [RFC8231] and H-PCE [RFC8685] features. This PCEP speaker will also include both the TLVs; in this case, a PCEP peer could falsely assume that the stateful H-PCE feature is also supported. On further PCEP message exchange, the stateful messages will not be propagated further (as described in this document), and a stateful H-PCE-based "parent" control of the LSP will not happen. A PCEP peer should be prepared for this eventuality as a part of normal procedures.¶
Apart from the basic error handling described in this document, an implementation could also use the enhanced error and notification mechanism for stateful H-PCE operations described in [PCE-ENHANCED-ERRORS]. Enhanced features such as error-behavior propagation, notification, and error-criticality level are further defined in [PCE-ENHANCED-ERRORS].¶
[RFC5623] describes a framework for applying the PCE-based architecture to interlayer (G)MPLS traffic engineering. The H-PCE stateful architecture with stateful P-PCE coordinating with the stateful C-PCEs of higher and lower layer is shown in Figure 2.¶
+----------+
| Parent |
/| PCE |
/ +----------+
/ / Stateful
/ / P-PCE
/ /
/ /
Stateful+-----+ / /
C-PCE | PCE |/ /
Hi | Hi | /
+-----+ /
+---+ +---+ / +---+ +---+
+ LSR +--+ LSR +........................+ LSR +--+ LSR +
+ H1 + + H2 + / + H3 + + H4 +
+---+ +---+\ +-----+/ /+---+ +---+
\ | PCE | /
\ | Lo | /
Stateful \ +-----+ /
C-PCE \ /
Lo \+---+ +---+/
+ LSR +--+ LSR +
+ L1 + + L2 +
+---+ +---+
All procedures described in Section 3 are also applicable to interlayer path setup, and therefore to separate domains.¶
It should be noted that if all the C-PCEs were to report all the LSPs in their domain, it could lead to scalability issues for the P-PCE. Thus, it is recommended to only report the LSPs that are involved in H-PCE -- i.e., the LSPs that are either delegated to the P-PCE or initiated by the P-PCE. Scalability considerations for PCEP as per [RFC8231] continue to apply for the PCEP session between child and parent PCE.¶
As described in Section 4.2 of [RFC6805], information about the content of child domains is not shared, for both scaling and confidentiality reasons. The child PCE could also conceal the path information during path computation. A C-PCE may replace a path segment with a path-key [RFC5520], effectively hiding the content of a segment of a path.¶
This document has no IANA actions.¶
Thanks to Manuela Scarella, Haomian Zheng, Sergio Marmo, Stefano Parodi, Giacomo Agostini, Jeff Tantsura, Rajan Rao, Adrian Farrel, and Haomian Zheng for their reviews and suggestions.¶
Thanks to Tal Mazrahi for the RTGDIR review, Paul Kyzivat for the GENART review, and Stephen Farrell for the SECDIR review.¶
Thanks to Barry Leiba, Martin Vigoureux, Benjamin Kaduk, and Roman Danyliw for the IESG review.¶