Huawei Technologies

Leela Palace Bangalore Karnataka 560008 INDIA udayasree.palle@huawei.com

Huawei Technologies

Leela Palace Bangalore Karnataka 560008 INDIA avantika.sushilkumar@huawei.com

Huawei Technologies

Leela Palace Bangalore Karnataka 560008 INDIA dhruv.ietf@gmail.com

Routing PCE Working Group The Path Computation Element (PCE) provides functions of path computation in support of traffic engineering in networks controlled by Multi-Protocol Label Switching (MPLS) and Generalized MPLS (GMPLS). This document provides extensions for the Path Computation Element Protocol (PCEP) to support multiple sources and destinations in a single path request.

specifies the Path Computation Element Communication Protocol (PCEP) for communications between a Path Computation Client (PCC) and a Path Computation Element (PCE), or between two PCEs, in compliance with . As per , a single Path Computation Request (PCReq) message may carry more than one path computation request. Each request is uniquely identified by a request-id number. In some scenarios (refer ), there is a need to send multiple requests with the same constraints and attributes to the PCE. Currently these requests are either sent in a separate PCReq messages or clubbed together in one (or more) PCReq messages. In either case, the constraints and attributes need to be encoded separately for each request even though they are exactly identical. Also note that, the PCE may choose to respond to each of the request independently in a separate Path Computation Reply (PCRep) messages or choose to bundle them in one (or more) PCRep messages. In some scenarios (refer ) one should wait for responses for all request before proceeding further. A mechanism to request path computation between multiple sources and destinations in a single path computation request would be helpful. Note that the scope of this work is point-to-point (P2P) path computation and is unrelated to MP2MP LSP ().

The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in .

The following terminology is used in this document. Label Switch Router. Multiprotocol Label Switching. Network Management System. Point-to-Point. Path Computation Client. Any client application requesting a path computation to be performed by a Path Computation Element. Path Computation Element. An entity (component, application, or network node) that is capable of computing a network path or route based on a network graph and applying computational constraints. Path Computation Element Communication Protocol.

Following key scenarios are identified where a mechanism to request path computation between multiple sources and destinations in a single path computation request would be helpful. describes the procedure for inter-domain path computation using Hierarchical PCE. In case of end to end path computation by Parent PCE, it needs to issue multiple path computation requests to child PCEs. In case of transit domain(s), the Parent PCE issues requests from each entry boundary node to each exit boundary node to the child PCE(s). Similarly for ingress domain, the Parent PCE issues requests from source to each exit boundary node, where as for egress domain, the Parent PCE issues requests from each entry boundary node to the destination. All requests to a particular child PCE, need to be encoded separately even though they are exactly identical (they have the same constraints and attributes). Also the Parent PCE needs to wait for responses for all requests before proceeding further. describes inter-layer path computation framework. In case of cooperating PCEs per layer, where each PCE has topology visibility restricted to its own layer and collaborate to compute an end-to-end path across layers. The higher layer PCE may need to issue multiple requests to lower layer PCE requesting paths from each entry boundary node to each exit boundary node. All these requests need to be encoded separately even though they are exactly identical (they have the same constraints and attributes). Also the higher layer PCE needs to wait for responses for all requests before proceeding further. describes a case where PCC is not necessarily an LSR, but for example, maybe a NMS or a planning tool etc. Such a PCC may issue multiple requests to PCE with identical constraints and attributes to select among the several source-destination pairs. In case where, for establishing a MP2MP TE LSP tunnel, multiple P2P path computation requests are sent to the PCE, one for each source-destination pair with identical constraints and attributes. In these scenarios, a mechanism to request path computation between multiple sources and destinations in a single path computation request would be helpful.

Consider the topology example mentioned in Section 4.6 of -

The following 11 individual requests are generated by parent PCE (PCE 5) - S-BN11; S-BN12; S-BN13; BN21-BN23; BN21-BN24; BN22-BN23; BN22-BN24; BN31-D; BN32-D; BN33-D; BN41-BN42; The above requests for each domain, need to be encoded separately even though they are exactly identical. A mechanism to request them together in a single path computation request would be helpful, in which case total 4 requests would be generated by parent PCE.

Following key requirements are identified for PCEP to enable multiple sources and destinations in a single path computation request: It MUST be possible for a single Path Computation Request to list multiple sources and destinations. It MUST be possible for a single Path Computation Reply to be sent for multiple sources and destinations. It MUST NOT be possible to set different constraints, traffic parameters, or quality-of-service requirements for different source and destination pair within a single computation request.

This document extends the existing END-POINTS object and by defining two new END-POINTS object types.

The END-POINTS object is used in a PCReq message to specify the source IP address and the destination IP address of the path for which a path computation is requested. This document extends the existing END-POINT object to support multiple sources and destinations in a single path request. Two new MP2MP END-POINTS objects for IPv4 and IPv6 are defined. The format of the MP2MP END-POINTS object body for IPv4 (Object-Type=5 (TBD)) is as follows:

The format of the MP2MP END-POINTS object body for IPv6 (Object-Type=6 (TBD)) is as follows:

The MP2MP END-POINTS object body has a variable length. These are multiples of 4 bytes for IPv4, and multiples of 16 bytes for IPv6, plus 4 bytes. On receiving MP2MP END-POINTS object, PCE computes m*n P2P paths, i.e, point to point path is computed between each combination of source and destination received in MP2MP END-POINTS object.

Since the END-POINTS object is not carried in the PCRep message (), the implementation supporting this extension SHOULD encode the source and the destination as the first and the last hop in the ERO. This is done to easily identify that which ERO corresponds to which source-destination pair.

As per , each request is uniquely identified by a request-id number. Since a single request is used for multiple sources and destinations sharing the same request-id number, along with request and response, the request-id number in RP object used in other PCEP messages (PCNtf, PCErr ...) applies to all sources and destinations (in the single request).

If PCE receives new END-POINTS type in path request and it understands the END-POINTS type, but the PCE is not capable of/support multiple sources and destinations in a single path request, the PCE MUST send a PCErr message with a PCEP-ERROR Object Error-Type = 4 (Not supported object) . The path computation request MUST then be cancelled. If the PCE does not understand the new END-POINTS type, then the PCE MUST send a PCErr message with a PCEP-ERROR Object Error-Type = 3 (Unknown object) .

The new END-POINTS type could be used to issue multiple computations together hence overloading the PCE. The PCE MUST allow for the use of policies to accept/reject such a request.

This document defines new END-POINTS types which does not add any new security concerns beyond those discussed in .

TBD.

TBD.

TBD.

TBD.

TBD.

TBD.

IANA assigns values to PCEP parameters in registries defined in . IANA is requested to make the following additional assignments.

Two new END-POINTS Object-Types are defined in this document. IANA is requested to make the following Object-Type allocations from the "PCEP Objects" sub-registry:

Thanks to Quintin Zhao for his suggestions.

Consider the domain 2 in , where 4 path requests need to be encoded (from 2 entry boundary nodes to 2 exit boundary nodes with the exact same constraints). Following figure illustrates the existing mechanism of carrying multiple requests in single PCReq message (as per ):

Combining multiple requests into a single request by using MP2MP END-POINTS object is illustrated below:

There is message size reduction of 64% in this example for just one domain.