PCEP Extension for Flow Specification

defines the Path Computation Element (PCE), a functional component capable of computing paths for use in traffic engineering networks. PCE was originally conceived for use in Multiprotocol Label Switching (MPLS) for Traffic Engineering (TE) networks to derive the routes of Label Switched Paths (LSPs). However, the scope of PCE was quickly extended to make it applicable to Generalized MPLS (GMPLS)-controlled networks, and more recent work has brought other traffic engineering technologies and planning applications into scope (for example, Segment Routing (SR) ). describes the Path Computation Element Communication Protocol (PCEP). PCEP defines the communication between a Path Computation Client (PCC) and a PCE, or between PCE and PCE, enabling computation of path for MPLS-TE LSPs. Stateful PCE specifies a set of extensions to PCEP to enable control of TE-LSPs by a PCE that retains state about the LSPs provisioned in the network (a stateful PCE). describes the setup, maintenance, and teardown of LSPs initiated by a stateful PCE without the need for local configuration on the PCC, thus allowing for a dynamic network that is centrally controlled. introduces the architecture for PCE as a central controller and describes how PCE can be viewed as a component that performs computation to place 'flows' within the network and decide how these flows are routed. The description of traffic flows by the combination of multiple Flow Specification Components and their dissemination as traffic flow specifications (Flow Specifications) is described for BGP in . A Flow Specification is comprised of traffic filtering rules and actions. The routers that receive a Flow Specification can classify received packets according to the traffic filtering rules and can direct packets based on the actions. When a PCE is used to initiate tunnels (such as TE-LSPs or SR paths) using PCEP, it is important that the head end of the tunnels understands what traffic to place on each tunnel. The data flows intended for a tunnel can be described using Flow Specification Components. When PCEP is in use for tunnel initiation it makes sense for that same protocol to be used to distribute the Flow Specification Components that describe what data is to flow on those tunnels. This document specifies a set of extensions to PCEP to support dissemination of Flow Specifications Components. For convenience we term the description of a traffic flow using Flow Specification Components as a "Flow Specification" and it must be understood that this is not the same as the same term used in since no action is explicitly included in the encoding. The extensions defined in this document include the creation, update, and withdrawal of Flow Specifications via PCEP, and can be applied to tunnels initiated by the PCE or to tunnels where control is delegated to the PCE by the PCC. Furthermore, a PCC requesting a new path can include Flow Specifications in the request to indicate the purpose of the tunnel allowing the PCE to factor this into the path computation. Flow Specifications are carried in TLVs within a new object called the FLOWSPEC object defined in this document. The flow filtering rules indicated by the Flow Specifications are mainly defined by BGP Flow Specifications.

This document uses the following terms defined in : PCC, PCE, PCEP Peer. The following term from is used frequently throughout this document: Flow Specification (FlowSpec): A Flow Specification is an n-tuple consisting of several matching criteria that can be applied to IP traffic, including filters and actions. Each FlowSpec consists of a set of filters and a set of actions. However, in the context of this document, no action is specified as part of the FlowSpec since the action "forward all matching traffic onto the associated path" is implicit. How an implementation decides how to filter traffic that matches a Flow Specification does not form part of this specification, but a flag is provided to indicate that the sender of a PCEP message that includes a Flow Specification is intended to be installed as a Longest Prefix Match route, or as a Flow Specification policy. This document uses the terms "stateful PCE" and "active PCE" as advocated in . 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 when, and only when, they appear in all capitals, as shown here.

In the PCE architecture there are five steps in the setup and use of LSPs: Decide which LSPs to set up. The decision may be made by a user, by a PCC, or by the PCE. There can be a number of triggers for this including user intervention and dynamic response to changes in traffic demands. Decide what properties to assign to an LSP. This can include bandwidth reservations, priorities, and DSCP (i.e., MPLS Traffic Class field). This function is also determined by user configuration or response to predicted or observed traffic demands. Decide what traffic to put on the LSP. This is effectively determining which traffic flows to assign to which LSPs, and practically, this is closely linked to the first two decisions listed above. Cause the LSP to be set up and modified to have the right characteristics. This will usually involve the PCE advising or instructing the PCC which will then signal the LSP across the network. Tell the head end what traffic to put on the LSP. This may happen after or at the same time as the LSP is set up. This step is the subject of this document.

There are three elements of procedure: A PCE and a PCC must be able to indicate whether or not they support the use of Flow Specifications. A PCE or PCC must be able to include Flow Specifications in PCEP messages with clear understanding of the applicability of those Flow Specifications in each case. This includes whether the use of such information is mandatory, constrained, or optional, and how overlapping Flow Specifications will be resolved. Flow Specification information/state must be synchronized between PCEP peers so that, on recovery, the peers have the same understanding of which Flow Specifications apply. The following subsections describe these points.

As with most PCEP capability advertisements, the ability to support Flow Specifications can be indicated in the PCEP OPEN message or in IGP PCE capability advertisements.

During PCEP session establishment, a PCC or PCE that supports the procedures described in this document announces this fact by including the "PCE FlowSpec Capability" TLV (described in ) in the OPEN Object carried in the PCEP Open message. The presence of the PCE FlowSpec Capability TLV in the OPEN Object in a PCE's OPEN message indicates that the PCE can distribute FlowSpecs to PCCs and can receive FlowSpecs in messages from PCCs. The presence of the PCE FlowSpec Capability TLV in the OPEN Object in a PCC's OPEN message indicates that the PCC supports the FlowSpec functionality described in this document. If either one of a pair of PCEP peers does not indicate support of the functionality described in this document by not including the PCE FlowSpec Capability TLV in the OPEN Object in its OPEN message, then the other peer MUST NOT include a FLOWSPEC object in any PCEP message sent to the peer that does not support the procedures. If a FLOWSPEC object is received when support has not been indicated, the receiver will respond with a PCErr message reporting the objects containing the FlowSpec as described in : that is, it will use 'Unknown Object' if it does not support this specification, and 'Not supported object' if it supports this specification but has not chosen to support FLOWSPEC objects on this PCEP session.

The ability to advertise support for PCEP and PCE features in IGP advertisements is provided for OSPF in and for IS-IS in . The mechanism uses the PCE Discovery TLV which has a PCE-CAP-FLAGS sub-TLV containing bit-flags each of which indicates support for a different feature. This document defines a new PCE-CAP-FLAGS sub-TLV bit, the FlowSpec Capable flag (bit number TBD1). Setting the bit indicates that an advertising PCE supports the procedures defined in this document. Note that while PCE FlowSpec Capability may be advertised during discovery, PCEP speakers that wish to use Flow Specification in PCEP MUST negotiate PCE FlowSpec Capability during PCEP session setup, as specified in . A PCC MAY initiate PCE FlowSpec Capability negotiation at PCEP session setup even if it did not receive any IGP PCE capability advertisement, and a PCEP peer that advertised support for FlowSpec in the IGP is not obliged to support these procedures on any given PCEP session.

This section describes the procedures to support Flow Specifications in PCEP messages. The primary purpose of distributing Flow Specification information is to allow a PCE to indicate to a PCC what traffic it should place on a path (such as an LSP or an SR path). This means that the Flow Specification may be included in: PCInitiate messages so that an active PCE can indicate the traffic to place on a path at the time that the PCE instantiates the path. PCUpd messages so that an active PCE can indicate or change the traffic to place on a path that has already been set up. PCRpt messages so that a PCC can report the traffic that the PCC plans to place on the path. PCReq messages so that a PCC can indicate what traffic it plans to place on a path at the time it requests the PCE to perform a computation in case that information aids the PCE in its work. PCRep messages so that a PCE that has been asked to compute a path can suggest which traffic could be placed on a path that a PCC may be about to set up. PCErr messages so that issues related to paths and the traffic they carry can be reported to the PCE by the PCC, and so that problems with other PCEP messages that carry Flow Specifications can be reported. To carry Flow Specifications in PCEP messages, this document defines a new PCEP object called the PCEP FLOWSPEC object. The object is OPTIONAL in the messages described above and MAY appear more than once in each message. The PCEP FLOWSPEC object carries zero or one Flow Filter TLV which describes a traffic flow. The inclusion of multiple PCEP FLOWSPEC objects allows multiple traffic flows to be placed on a single path. Once a PCE and PCC have established that they can both support the use of Flow Specifications in PCEP messages, such information may be exchanged at any time for new or existing paths. The application and prioritization of Flow Specifications is described in . As per , any attributes of the path received from a PCE are subject to PCC's local policy. This holds good for the Flow Specifications as well.

The Flow Specifications are carried along with the LSP State information as per making the Flow Specifications part of the LSP database (LSP-DB). Thus, the synchronization of the Flow Specification information is done as part of LSP-DB synchronization. This may be achieved using normal state synchronization procedures as described in or enhanced state synchronization procedures as defined in . The approach selected will be implementation and deployment specific and will depend on issues such as how the databases are constructed and what level of synchronization support is needed.

The PCE-FLOWSPEC-CAPABILITY TLV is an optional TLV that can be carried in the OPEN Object to exchange PCE FlowSpec capabilities of the PCEP speakers. The format of the PCE-FLOWSPEC-CAPABILITY TLV follows the format of all PCEP TLVs as defined in and is shown in .

The type of the PCE-FLOWSPEC-CAPABILITY TLV is TBD2 and it has a fixed length of 2 octets. The Value field is set to default value 0. The two bytes of padding MUST be set to zero and ignored on receipt. The inclusion of this TLV in an OPEN object indicates that the sender can perform FlowSpec handling as defined in this document.

The PCEP FLOWSPEC object defined in this document is compliant with the PCEP object format defined in . It is OPTIONAL in the PCReq, PCRep, PCErr, PCInitiate, PCRpt, and PCUpd messages and MAY be present zero, one, or more times. Each instance of the object specifies a traffic flow. The PCEP FLOWSPEC object carries a FlowSpec filter rule encoded in a TLV (as defined in ). The FLOWSPEC Object-Class is TBD3 (to be assigned by IANA). The FLOWSPEC Object-Type is 1. The format of the body of the PCEP FLOWSPEC object is shown in

FS-ID (32-bits): A PCEP-specific identifier for the FlowSpec information. A PCE or PCC creates an FS-ID for each FlowSpec that it originates, and the value is unique within the scope of that PCE or PCC and is constant for the lifetime of a PCEP session. All subsequent PCEP messages can identify the FlowSpec using the FS-ID. The values 0 and 0xFFFFFFFF are reserved and MUST NOT be used. AFI (16-bits): Address Family Identifier as used in BGP (AFI=1 for IPv4 or VPNv4, AFI=2 for IPv6 and VPNv6 as per as per ). Reserved (8-bits): MUST be set to zero on transmission and ignored on receipt. Flags (8-bits): Two flags are currently assigned - R bit: The Remove bit is set when a PCEP FLOWSPEC object is included in a PCEP message to indicate removal of the Flow Specification from the associated tunnel. If the bit is clear, the Flow Specification is being added or modified. L bit: The Longest Prefix Match (LPM) bit is set to indicate that the Flow Specification is to be installed as a route subject to longest prefix match forwarding. If the bit is clear, the Flow Specification described by the Flow Filter TLV (see ) is to be installed as a Flow Specification. If the bit is set, only Flow Specifications that describe IPv4 or IPv6 destinations are meaningful in the Flow Filter TLV. If the L is set and the receiver does not support the use of Flow Specifications that are present in the Flow Filter TLV for the installation of a route subject to longest prefix match forwarding, then the PCEP peer MUST respond with a PCErr message with error-type TBD8 (FlowSpec Error) and error-value 5 (Unsupported LPM Route). Unassigned bits MUST be set to zero on transmission and ignored on receipt. If the PCEP speaker receives a message with R bit set in the FLOWSPEC object and the Flow Specification identified with a FS-ID does not exist, it MUST generate a PCErr with Error-type TBD8 (FlowSpec Error), error-value 4 (Unknown FlowSpec). If the PCEP speaker does not understand or support the AFI in the FLOWSPEC message, the PCEP peer MUST respond with a PCErr message with error-type TBD8 (FlowSpec Error), error-value 2 (Malformed FlowSpec). The following TLVs can be used in the FLOWSPEC object: Speaker Entity Identifier TLV: As specified in , SPEAKER-ENTITY-ID TLV encodes a unique identifier for the node that does not change during the lifetime of the PCEP speaker. This is used to uniquely identify the FlowSpec originator and thus used in conjunction with FS-ID to uniquely identify the FlowSpec information. This TLV MUST be included. If the TLV is missing, the PCEP peer MUST respond with a PCErr message with error-type TBD8 (FlowSpec Error), error-value 2 (Malformed FlowSpec). Flow Filter TLV (variable): One TLV MAY be included. The Flow Filter TLV is OPTIONAL when the R bit is set. The TLV MUST be present when the R bit is clear. If the TLV is missing when the R bit is clear, the PCEP peer MUST respond with a PCErr message with error-type TBD8 (FlowSpec Error) and error-value 2 (Malformed FlowSpec).

A new PCEP TLV is defined to convey Flow Specification filtering rules that specify what traffic is carried on a path. The TLV follows the format of all PCEP TLVs as defined in . The Type field values come from the codepoint space for PCEP TLVs and has the value TBD4. The Value field contains one or more sub-TLVs (the Flow Specification TLVs) as defined in . Only one Flow Filter TLV can be present and represents the complete definition of a Flow Specification for traffic to be placed on the tunnel. This tunnel is indicated by the PCEP message in which the PCEP FLOWSPEC object is carried. The set of Flow Specification TLVs in a single instance of a Flow Filter TLV are combined to indicate the specific Flow Specification. Further Flow Specifications can be included in a PCEP message by including additional FLOWSPEC objects.

The Flow Filter TLV carries one or more Flow Specification TLV. The Flow Specification TLV follows the format of all PCEP TLVs as defined in . However, the Type values are selected from a separate IANA registry (see ) rather than from the common PCEP TLV registry. Type values are chosen so that there can be commonality with Flow Specifications defined for use with BGP . This is possible because the BGP Flow Spec encoding uses a single octet to encode the type where as PCEP uses two octets. Thus the space of values for the Type field is partitioned as shown in .

is the reference for the registry "Flow Spec Component Types" and defines the allocations it contains. requested for another registry "Flow Spec IPv6 Component Types" and requested initial allocations in it. If the AFI (in the FLOWSPEC object) is set to IPv4, the range 1..255 is as per "Flow Spec Component Types" ; if the AFI is set to IPv6, the range 1..255 is as per "Flow Spec IPv6 Component Types" . When future BGP specifications (such as ) make further allocations to the aforementioned registries, they are also inherited for PCEP usage. The content of the Value field in each TLV is specific to the type/AFI and describes the parameters of the Flow Specification. The definition of the format of many of these Value fields is inherited from BGP specifications. Specifically, the inheritance is from and , but may also be inherited from future BGP specifications. When multiple Flow Specification TLVs are present in a single Flow Filter TLV they are combined to produce a more detailed specification of a flow. For examples and rules about how this is achieved, see . An implementation that receives a PCEP message carrying a Flow Specification TLV with a type value that it does not recognize or does not support MUST respond with a PCErr message with error-type TBD8 (FlowSpec Error), error-value 1 (Unsupported FlowSpec) and MUST NOT install the Flow Specification. When used in other protocols (such as BGP), these Flow Specifications are also associated with actions to indicate how traffic matching the Flow Specification should be treated. In PCEP, however, the only action is to associate the traffic with a tunnel and to forward matching traffic onto that path, so no encoding of an action is needed. describes how overlapping Flow Specifications are prioritized and handled. All Flow Specification TLVs with Types in the range 1 to 255 have Values defined for use in BGP (for example, in , , and ) and are set using the BGP encoding, but without the type octet (the relevant information is in the Type field of the TLV). The Value field is padded with trailing zeros to achieve 4-byte alignment. This document defines following new types -

To allow identification of a VPN in PCEP via a Route Distinguisher (RD) , a new TLV - ROUTE-DISTINGUISHER TLV is defined in this document. A Flow Specification TLV with Type TBD5 (ROUTE-DISTINGUISHER TLV) carries an RD Value, used to identify that other flow filter information (for example, an IPv4 destination prefix) is associated with a specific VPN identified by the RD. See for further discussion of VPN identification.

The format of RD is as per . Although it may be possible to describe a multicast Flow Specification from the combination of other Flow Specification TLVs with specific values, it is more convenient to use a dedicated Flow Specification TLV. Flow Specification TLVs with Type values TBD6 and TBD7 are used to identify a multicast flow for IPv4 and IPv6 respectively. The Value field is encoded as shown in .

The address fields and address mask lengths of the two Multicast Flow Specification TLVs contain source and group prefixes for matching against packet flows noting that the two address fields are 32 bits for an IPv4 Multicast Flow and 128 bits for an IPv6 Multicast Flow. The Reserved field MUST be set to zero and ignored on receipt. Two bit flags (S and G) are defined to describe the multicast wildcarding in use. If the S bit is set, then source wildcarding is in use and the values in the Source Mask Length and Source Address fields MUST be ignored. If the G bit is set, then group wildcarding is in use and the values in the Group Mask Length and Group multicast Address fields MUST be ignored. The G bit MUST NOT be set unless the S bit is also set: if a Multicast Flow Specification TLV is received with S bit = 0 and G bit = 1 the receiver MUST respond with a PCErr with Error-type TBD8 (FlowSpec Error) and error-value 2 (Malformed FlowSpec). The three multicast mappings may be achieved as follows: (S, G) - S bit = 0, G bit = 0, the Source Address and Group multicast Address prefixes are both used to define the multicast flow. (*, G) - S bit = 1, G bit = 0, the Group multicast Address prefix, but the Source Address prefix is ignored. (*, *) = S bit = 1, G bit = 1, the Source Address and Group multicast Address prefixes are both ignored.

This section outlines some specific detailed procedures for using the protocol extensions defined in this document.

The default behavior is that no Flow Specification is applied to a tunnel. That is, the default is that the FLOWSPEC object is not used as is the case in all systems before the implementation of this specification. In this case, it is a local matter (such as through configuration) how tunnel head ends are instructed what traffic to place on a tunnel. describes how receivers respond when they see unknown PCEP objects.

Flow Specifications may be represented by a single Flow Specification TLV or may require a more complex description using multiple Flow Specification TLVs. For example, a flow indicated by a source-destination pair of IPv6 addresses would be described by the combination of Destination IPv6 Prefix and Source IPv6 Prefix Flow Specification TLVs.

A PCE may want to modify a Flow Specification associated with a tunnel, or a PCC may want to report a change to the Flow Specification it is using with a tunnel. It is important that the specific Flow Specification is identified so that it is clear that this is a modification of an existing flow and not the addition of a new flow as described in . The FS-ID field of the PCEP FLOWSPEC object is used to identify a specific Flow Specification. When modifying a Flow Specification, all Flow Specification TLVs for the intended specification of the flow MUST be included in the PCEP FLOWSPEC object and the FS-ID MUST be retained from the previous description of the flow.

It is possible that multiple flows will be place on a single tunnel. In some cases it is possible to to define these within a single PCEP FLOWSPEC object: for example, two Destination IPv4 Prefix TLVs could be included to indicate that packets matching either prefix are acceptable. PCEP would consider this as a single Flow Specification identified by a single FS-ID. In other scenarios the use of multiple Flow Specification TLVs would be confusing. For example, if flows from A to B and from C to D are to be included then using two Source IPv4 Prefix TLVs and two Destination IPv4 Prefix TLVs would be confusing (are flows from A to D included?). In these cases, each Flow Specification is carried in its own PCEP FLOWSPEC object with multiple objects present on a single PCEP message. Use of separate objects also allows easier removal and modification of Flow Specifications.

The Remove bit in the PCEP FLOWSPEC object is left clear when a Flow Specification is being added or modified. To remove a Flow Specification, a PCEP FLOWSPEC object is included with the FS-ID matching the one being removed, and the R bit set to indicate removal. In this case it is not necessary to include any Flow Specification TLVs. If the R bit is set and Flow Specification TLVs are present, an implementation MAY ignore them. If the implementation checks the Flow Specification TLVs against those recorded for the FS-ID of the Flow Specification being removed and finds a mismatch, the Flow Specification MUST still be removed and the implementation SHOULD record a local exception or log.

VPN instances are identified in BGP using Route Distinguishers (RDs) . These values are not normally considered to have any meaning outside of the network, and they are not encoded in data packets belonging to the VPNs. However, RDs provide a useful way of identifying VPN instances and are often manually or automatically assigned to VPNs as they are provisioned. Thus the RD provides a useful way to indicate that traffic for a particular VPN should be placed on a given tunnel. The tunnel head end will need to interpret this Flow Specification not as a filter on the fields of data packets, but using the other mechanisms that it already uses to identify VPN traffic. This could be based on the incoming port (for port-based VPNs) or may leverage knowledge of the VRF that is in use for the traffic.

Flow specifications can overlap. For example, two different flow specifications may be identical except for the length of the prefix in the destination address. In these cases the PCC must determine how to prioritize the flow specifications so as to know to which path to assign packets that match both flow specifications. That is, the PCC must assign a precedence to the flow specifications so that it checks each incoming packet for a match in a predictable order. The processing of BGP Flow Specifications is described in . Section 5.1 of that document explains the order of traffic filtering rules to be executed by an implementation of that specification. PCCs MUST apply the same ordering rules as defined in . Furthermore, it is possible that Flow Specifications will be distributed by BGP as well as by PCEP as described in this document. In such cases implementations supporting both approaches MUST apply the prioritization and ordering rules as set out in regardless of which protocol distributed the Flow Specifications, however implementations MAY provide a configuration control to allow one protocol to take precedence over the other as this may be particularly useful if the Flow Specification make identical matches on traffic but have different actions. It is RECOMMENDED that when two Flow Specifications distributed by different protocols overlap, and especially when one acts to replace another, that a message be logged for the operator to understand the behaviour. of this document covers manageability considerations relevant to the prioritized ordering of flow specifications. An implementation that receives a PCEP message carrying a Flow Specification that it cannot resolve against other Flow Specifications already installed MUST respond with a PCErr message with error-type TBD8 (FlowSpec Error), error-value 3 (Unresolvable Conflict) and MUST NOT install the Flow Specification.

This section describes the format of messages that contain FLOWSPEC objects. The only difference to previous message formats is the inclusion of that object. The figures in this section use the notation defined in . The FLOWSPEC object is OPTIONAL and MAY be carried in the PCEP messages. The PCInitiate message is defined in and updated as below:

::=

Where: ::= [] ::= ( | ) ::= [] [] [] Where: ::= [] ]]>

The PCUpd message is defined in and updated as below:

::=

Where: ::= [] ::= [] Where: ::=

::= [] ]]>

The PCRpt message is defined in and updated as below:

::=

Where: ::= [] ::= [] [] Where: ::= [

]

::= [] ]]>

The PCReq message is defined in and updated in , it is further updated below for flow specification:

::= [] Where: ::= [] ::= [] ::=

[] [] [] [] [[]] [] [] [] Where: ::= [] ]]>

The PCRep message is defined in and updated in , it is further updated below for flow specification:

::=

Where: ::=[] ::= [] [] [] [] [] Where: ::= [] ]]>

IANA maintains the "Path Computation Element Protocol (PCEP) Numbers" registry. This document requests IANA actions to allocate code points for the protocol elements defined in this document.

Each PCEP object has an Object-Class and an Object-Type. IANA maintains a subregistry called "PCEP Objects". IANA is requested to make an assignment from this subregistry as follows:

This document requests that a new sub-registry, named "FLOWSPEC Object Flag Field", is created within the "Path Computation Element Protocol (PCEP) Numbers" registry to manage the Flag field of the FLOWSPEC object. New values are to be assigned by Standards Action . Each bit should be tracked with the following qualities: Bit number (counting from bit 0 as the most significant bit) Capability description Defining RFC The initial population of this registry is as follows:

IANA maintains a subregistry called "PCEP TLV Type Indicators". IANA is requested to make an assignment from this subregistry as follows:

IANA is requested to create a new subregistry call the "PCEP Flow Specification TLV Type Indicators" registry. Allocations from this registry are to be made according to the following assignment policies :

IANA is requested to pre-populate this registry with values defined in this document as follows, taking the new values from the range 256 to 64506:

IANA maintains a subregistry called "PCEP-ERROR Object Error Types and Values". Entries in this subregistry are described by Error-Type and Error-value. IANA is requested to make the following assignment from this subregistry:

IANA maintains a subregistry called "Open Shortest Path First v2 (OSPFv2) Parameters" with a sub-registry called "Path Computation Element (PCE) Capability Flags". IANA is requested to assign a new capability bit from this registry as follows:

[NOTE TO RFC EDITOR : This whole section and the reference to RFC 7942 is to be removed before publication as an RFC] This section records the status of known implementations of the protocol defined by this specification at the time of posting of this Internet-Draft, and is based on a proposal described in . The description of implementations in this section is intended to assist the IETF in its decision processes in progressing drafts to RFCs. Please note that the listing of any individual implementation here does not imply endorsement by the IETF. Furthermore, no effort has been spent to verify the information presented here that was supplied by IETF contributors. This is not intended as, and must not be construed to be, a catalog of available implementations or their features. Readers are advised to note that other implementations may exist. According to , "this will allow reviewers and working groups to assign due consideration to documents that have the benefit of running code, which may serve as evidence of valuable experimentation and feedback that have made the implemented protocols more mature. It is up to the individual working groups to use this information as they see fit". At the time of posting the -04 version of this document, there are no known implementations of this mechanism. It is believed that two vendors are considering prototype implementations, but these plans are too vague to make any further assertions.

We may assume that a system that utilizes a remote PCE is subject to a number of vulnerabilities that could allow spurious LSPs or SR paths to be established or that could result in existing paths being modified or torn down. Such systems, therefore, apply security considerations as described in , , and . The description of Flow Specifications associated with paths set up or controlled by a PCE add a further detail that could be attacked without tearing down LSPs or SR paths, but causing traffic to be misrouted within the network. Therefore, the use of the security mechanisms for PCEP referenced above is important. Visibility into the information carried in PCEP does not have direct privacy concerns for end-users' data, however, knowledge of how data is routed in a network may make that data more vulnerable. Of course, the ability to interfere with the way data is routed also makes the data more vulnerable. Furthermore, knowledge of the connected end-points (such as multicast receivers or VPN sites) is usually considered private customer information. Therefore, implementations or deployments concerned with protecting privacy MUST apply the mechanisms described in the documents referenced above. Experience with Flow Specifications in BGP systems indicates that they can become complex and that the overlap of Flow Specifications installed in different orders can lead to unexpected results. Although this is not directly a security issue per se, the confusion and unexpected forwarding behavior may be engineered or exploited by an attacker. Therefore, implementers and operators SHOULD pay careful attention to the Manageability Considerations described in .

The feature introduced by this document enables operational manageability of networks operated in conjunction with a PCE and using PCEP. Without this feature, but in the case of a stateful active PCE or with PCE-initiated services, additional manual configuration is needed to tell the head-ends what traffic to place on the network services (LSPs, SR paths, etc.). This section follows the advice and guidance of .

Experience with flow specification in BGP suggests that there can be a lot of complexity when two or more flow specifications overlap. This can arise, for example, with addresses indicated using prefixes, and could cause confusion about what traffic should be placed on which path. Unlike the behavior in a distributed routing system, it is not important that each head-end implementation applies the same rules to disambiguate overlapping Flow Specifications, but it is important that: A network operator can easily find out what traffic is being placed on which path and why. This will facilitate analysis of the network and diagnosis of faults. A PCE is able to correctly predict the effect of instructions it gives to a PCC. To that end, a PCC MUST enable an operator to view the the Flow Specifications that it has installed, and these MUST be presented in order of precedence such that when two Flow Specifications overlap, the one that will be serviced with higher precedence is presented to the operator first. A discussion of precedence ordering for flow specifications is found in .

Support for the function described in this document implies that a functional element that is capable of requesting a PCE to compute and control a path is also able to configure the specification of what traffic should be placed on that path. Where there is a human involved in this action, configuration of the Flow Specification must be available through an interface (such as a graphical user interface or a command line interface). Where a distinct software component (i.e., one not co-implemented with the PCE) is used, a protocol mechanism will be required that could be PCEP itself or could be a data model such as extensions to the YANG model for requesting path computation . Implementations MAY be constructed with a configurable switch to say whether they support the functions defined in this document. Otherwise, such implementations MUST indicate that they support the function as described in . If an implementation supports configurable support of this function, that support MAY be configurable per peer or once for the whole implementation. As mentioned in , a PCE implementation SHOULD provide a mechanism to configure variations in the precedence ordering of Flow Specifications per PCC.

The YANG model in can be used to model and monitor PCEP states and messages. To make that YANG model useful for the extensions described in this document, it will need to be augmented to cover the new protocol elements. Similarly, as noted in , the YANG model defined in could be extended to allow specification of Flow Specifications. Finally, as mentioned in , a PCC implementation SHOULD provide a mechanism to allow an operator to read the Flow Specifications from a PCC and to understand in what order they will be executed. This could be achieved using a new YANG model.

The extensions defined in this document do not require any additional liveness detection and monitoring support. See and for more information.

The chief element of operation that needs to be verified (in addition to the operation of the protocol elements as described in ) is the installation, precedence, and correct operation of the Flow Specifications at a PCC. In addition to the YANG model for reading Flow Specifications described in , tools may be needed to inject Operations and Management (OAM) traffic at the PCC that matches specific criteria so that it can be monitored as traveling along the desired path. Such tools are outside the scope of this document.

This document places no requirements on other protocols or components.

The use of the features described in this document clearly have an important impact on network traffic since they cause traffic to be routed on specific paths in the network. However, in practice, these changes make no direct changes to the network operation because traffic is already placed on those paths using some pre-existing configuration mechanism. Thus, the significant change is the reduction in mechanisms that have to be applied, rather than a change to how the traffic is passed through the network.

No other manageability considerations are known at this time.

Thanks to Julian Lucek, Sudhir Cheruathur, Olivier Dugeon, Jayant Agarwal, Jeffrey Zhang, Acee Lindem, and Vishnu Payam Beeram for useful discussions.