Path Computation Element communication
Protocol (PCEP) extension for signaling LSP diversity constraintOrangestephane.litkowski@orange.comCisco Systems, Inc.2000 Innovation DriveKanataOntarioK2K 3E8Canadamsiva@cisco.comJuniper Networkscbarth@juniper.netHuaweiDivyashree Techno Park, WhitefieldBangaloreKA560066Indiadhruv.ietf@gmail.com
Routing
PCE Working Group
This document introduces a simple mechanism to associate
a group of Label Switched Paths (LSPs) via an extension to the Path Computation
Element (PCE) Communication Protocol (PCEP) with the purpose of computing diverse paths for those LSPs.
The proposed extension allows a Path Computation Client (PCC) to advertise to a PCE that a particular LSP belongs to a disjoint-group,
thus the PCE knows that LSPs in the same group needs to be disjoint from each other. describes the Path Computation Element communication
Protocol (PCEP) which enables the communication between a Path
Computation Client (PCC) and a Path Control Element (PCE), or between
two PCEs based on the PCE architecture .
PCEP Extensions for Stateful PCE Model
describes a set of extensions to PCEP to enable active control of
MPLS-TE and GMPLS tunnels.
describes the setup and teardown of PCE-initiated LSPs under the
active stateful PCE model, without the need for local configuration
on the PCC, thus allowing for a dynamic network.
introduces a generic
mechanism to create a grouping of LSPs which can then be used to
define associations between a set of LSPs and a set of attributes (such
as configuration parameters or behaviors) and is equally applicable
to the active and passive modes of a stateful PCE or a stateless PCE .This document specifies a PCEP extension to signal that a particular group of
LSPs should use diverse paths including the requested type of diversity.
A PCC can use this extension to signal to a PCE that a particular LSP belongs to a disjoint-group.
When a PCE receives LSP states belonging to the same disjoint-group from some PCCs, the PCE should ensure that the LSPs within the group are disjoint from each other.
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.The following terminology is used in this document.Label Switch Router.Multiprotocol Label Switching.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.Shared Risk Link Group.Path diversity is a very common use case in today's IP/MPLS networks especially for layer 2 transport over MPLS.
A customer may request that the operator provide two end-to-end disjoint paths across the IP/MPLS core.
The customer may use those paths as primary/backup or active/active.
Different level of disjointness may be offered:
Link disjointness: the paths of the associated LSPs should transit different
links (but may use common nodes or different links that may have some shared
fate).
Node disjointness: the paths of the associated LSPs should transit different
nodes (but may use different links that may have some shared fate).
SRLG disjointness: the paths of the associated LSPs should transit different
links that do not share fate (but may use common transit nodes).
Node+SRLG disjointness: the paths of the associated LSPs should transit
different links that do not have any common shared fate and should transit
different nodes.
The associated LSPs may originate from the same or from different head-end(s) and may terminate at the same or different tail-end(s).
In the figure above, consider that the customer wants to have two disjoint paths between CE1/CE2 and CE3/CE4. From an IP/MPLS network point view, in this example, the CEs are connected to different PEs to maximize their disjointness.
When LSPs originate from different head-ends, distributed computation
of diverse paths can be difficult. Whereas, computation via a centralized PCE ensures
path disjointness correctness and simplicity.
defines a mechanism for the synchronization of a
set of path computation requests by using the SVEC (Synchronization
VECtor) object, that specifies the list of synchronized requests that
can either be dependent or independent. The SVEC object identify the
relationship between the set of path computation requests, identified
by 'Request-ID-number' in RP (Request Parameters) object.
further clarified the use of the SVEC list for synchronized path
computations when computing dependent requests as well as described a
number of usage scenarios for SVEC lists within single-domain and
multi-domain environments.
The SVEC object includes a Flags field that indicates the potential
dependency between the set of path computation request in a similar
way as the Flags field in the TLVs defined in this document. The
path computation request in the PCReq message MAY use both SVEC object to
identify the related path computation request as well as to identify
the diversity association group. The PCE MUST try to find a path that
meets both the constraints. It is possible that the diversity set in
the association group is different from the one in SVEC object, this
might be true for the same LSP as well. The PCE would consider both
the objects as per the processing rules and aim to find a path that
meets both these constraints. In case no such path is possible (or the
constraints are incompatible), the PCE MUST send a path computation
reply (PCRep) with NO-PATH object indicating path computation failure
as per .
The PCEP extension for stateful PCE defined
new PCEP messages - PCRpt, PCUpd and PCInitiate . These messages uses
PLSP-ID in the LSP object for identification. Moreover to allow diversity
between LSPs originating from different PCCs, the generic mechanism to
create a grouping of LSPs is described in
(that is equally applicable to the active and passive modes of a stateful PCE).
Using PCEP, the PCC could indicate that the disjoint path computation is required, such indication should include disjointness parameters such as the type of
disjointness, the disjoint group identifiers, and any customization parameters according to
the configured local policy. As mentioned previously, the extension described in is well suited to associate a set
of LSPs with a particular disjoint-group.The management of the disjoint group-ids will be a key point for the operator as the Association ID field is limited to 65535.
The local configuration of IPv4/IPv6 association source, or Global Association Source/Extended Association ID should allow to overcome this limitation as described in .
When a PCC or PCE initiates all the LSPs in a particular disjoint-group, it can set the IPv4/IPv6 association source as one of its own IP address.
When disjoint LSPs are initiated from different head-ends, association source could be the PCE address or any other unique value to identify the disjoint association group.
Using the disjoint-group within a PCEP messages may have two purpose:
Information: in case the PCE is performing the path computation, it may communicate to the PCC the disjoint parameters.Configuration: in case the PCC are configured with disjoint requirements, these are communicated to the PCE.As per , LSPs
are associated with other LSPs with which they interact by adding
them to a common association group. The Association parameters, as described in as the combination of the mandatory fields Association type, Association
ID and Association Source in the ASSOCIATION object, that uniquely identify
the association group, uniquely identify the disjoint group. If the optional TLVs - Global Association
Source or Extended Association ID are included, then they are
included in combination with mandatory fields to uniquely identifying
the association group.
This document defines a new Association type,
based on the generic Association object -
Association type = TBD1 ("Disjointness
Association Type") for Disjoint Association Group (DAG). specify the mechanism for the capability
advertisement of the association types supported by a PCEP speaker by defining a ASSOC-Type-List TLV to be carried within an OPEN object.
This capability exchange for the association type described in this document (i.e. Disjointness Association Type) MUST be done before using the disjointness association.
Thus the PCEP speaker MUST include the Disjointness Association Type (TBD1) in the ASSOC-Type-List TLV
before using the disjoint association group (DAG) in the PCEP messages.This association type is considered to be both dynamic and operator-configured in nature. The association group could be
created by the operator manually on the PCEP peers and the LSPs belonging to
this associations is conveyed via PCEP messages to the PCEP peer; or the association group could be created dynamically by the PCEP speaker and both the association group information and the LSPs belonging to the association group is conveyed to the PCEP peer. The Operator-configured
Association Range MUST be set for this association-type to mark a range of association identifiers that
are used for operator-configured associations to avoid any
association identifier clash within the scope of the association
source. (Refer .)A disjoint group can have two or more LSPs. But a PCE may be limited in how many LSPs it can take into account when computing disjointness.
If a PCE receives more LSPs in the group than it can handle in its computation algorithm, it SHOULD apply disjointness computation to only a subset of LSPs in the group. The subset of disjoint LSPs will be decided by PCE as a local matter.
Local polices on the PCC or PCE MAY define the computational behavior for the other LSPs
in the group. For example, the PCE may provide no path, a shortest path, or a constrained path
based on relaxing disjointness, etc.Associating a particular LSP to multiple disjoint groups is authorized from a protocol perspective, however there is no insurance that the PCE will be able to compute properly the multi-disjointness constraint.
The disjoint group MUST carry the following TLV:
DISJOINTNESS-CONFIGURATION-TLV: Used to communicate some disjointness configuration parameters.
In addition, the disjoint group MAY carry the following TLV:
DISJOINTNESS-STATUS-TLV: Used to communicate the status of the computed disjointness. This is applicable for messages from PCE to PCC (PCUpd, PCInitiate or PCRep message).VENDOR-INFORMATION-TLV: Used to communicate arbitrary vendor specific behavioral
information, described in .
The DISJOINTNESS-CONFIGURATION-TLV is shown in the following figure:
Type: TBD2. Length: Fixed value of 4 bytes.
Flags:
L (Link diverse) bit: when set, this indicates that the
computed paths within the disjoint group MUST NOT have any link in common.N (Node diverse) bit: when set, this indicates that the
computed paths within the disjoint group MUST NOT have any node in common.S (SRLG diverse) bit: when set, this indicates that the
computed paths within the disjoint group MUST NOT share any SRLG (Shared Risk Link
Group).P (Shortest path) bit: when set, this indicates that the computed path of the LSP SHOULD satisfies all constraints and objective functions first without considering the diversity constraint.
This means that an LSP with P flag set should be placed as if the disjointness constraint has not been configured, while the other LSP in the association with P flag unset should be placed by taking into account the disjointness constraint. Setting P flag changes the relationship between LSPs to a unidirectional relationship (LSP 1 with P=0 depends of LSP 2 with P=1, but LSP 2 with P=1 does not depend of LSP 1 with P=0).T (Strict disjointness) bit: when set, if disjoint paths cannot be found, PCE should return no path for LSPs that could not be be disjoint. When unset, PCE is allowed to relax disjointness by using either applying a requested objective function or any other behavior if no objective function is requested (e.g.: using a lower disjoint type (link instead of node) or relaxing disjointness constraint at all).
If a PCEP speaker receives a disjoint-group without DISJOINTNESS-CONFIGURATION-TLV, it SHOULD reply with a PCErr Error-type=6 (Mandatory Object missing) and Error-value=TBD7 (DISJOINTNESS-CONFIGURATION-TLV missing).
The DISJOINTNESS-STATUS-TLV uses the same format as the DISJOINTNESS-CONFIGURATION-TLV with a different type TBD3 (in TLV):
Any new flag defined for the DISJOINTNESS-CONFIGURATION-TLV is be automatically applicable to the DISJOINTNESS-STATUS-TLV.
An objective function (OF) MAY be applied to the disjointness computation
to drive the PCE computation behavior. In this case, the OF-List TLV
(defined in () is used as an optional TLV in the Association
Group Object. The PCEP OF-List TLV allow multiple OF-Codes inside the
TLV, a sender SHOULD include a single OF-Code in the OF-List TLV when
included in the Association Group, and the receiver MUST consider the
first OF-code only and ignore others if included.
To minimize the common shared resources (Node, Link or SRLG) between
a set of paths during path computation three new OF codes are
proposed:
MSLMinimize the number of shared (common) Links.TBD4Find a set of paths such that it passes through the least number of shared (common) links.MSSMinimize the number of shared (common) SRLGs.TBD5Find a set of paths such that it passes through the least number of shared (common) SRLGs.MSNMinimize the number of shared (common) Nodes.TBD6Find a set of paths such that it passes through the least number of shared (common) nodes. uses SVEC diversity flag for node, link or SRLG to describe the potential disjointness between the set of path computation requests used in PCEP protocol.
This document defines three new OF codes to maximize diversity as much as possible, in other words, minimize the common shared resources (Node,Link or SRLG) between a set of paths.
It may be interesting to note that the diversity flags in the SVEC object and OF for diversity can be used together. Some example of usage are listed below -
SVEC object with node-diverse bit=1 - ensure full node-diversity.
SVEC object with node-diverse bit=1 and OF=MSS - full node diverse with as much as SRLG-diversity as possible.
SVEC object with domain-diverse bit=1;link diverse bit=1 and OF=MSS - full domain and node diverse path with as much as SRLG-diversity as possible.
SVEC object with node-diverse bit=1 and OF=MSN - ensure full node-diversity.
As mentioned in , the P-flag (when set) indicates that the computed path of the LSP SHOULD satisfies all constraints and objective functions first without considering the diversity constraint. This could be required in some primary/backup scenarios where the primary path should use the more optimal path available (taking into account the other constraints).
When disjointness is computed, it is important for the algorithm to know that it should try to optimize the path of one or more LSPs in the disjoint group (for instance the primary path) while other paths are allowed to be longer (compared to a similar path without the disjointness constraint).
Without such a hint, the disjointness algorithm may set a path for all LSPs that may not completely fulfill the customer requirement.
In the figure above, a customer has two dual homed sites (CE1/CE3 and CE2/CE4). Consider, this customer wants two disjoint paths between the two sites.
Due to physical meshing, the customer wants to use CE1 and CE2 as primary ( and CE3 and CE4 are hosted in a remote site for redundancy purpose).
Without any hint (constraint) provided, the PCE may compute the two disjoint LSPs together, leading to PE1->PE2 using a path PE1->R1->R2->PE2 and PE3->PE4 using PE3->R3->R4->PE4.
In this case, even if the disjointness constraint is fulfilled, the path from PE1 to PE2 does not use the best optimal path available in the network (RTD may be higher): the customer requirement is thus not completely fulfilled.
The usage of the P-Flag allows the PCE to know that a particular LSP should be tied to the best path as if the disjointness constraint was not requested.In our example, if the P-Flag is set to the LSP PE1->PE2, the PCE should use the path PE1->R1->R3->R4->R2->PE2 for this LSP, while the other LSP should be disjoint from this path.
The second LSP will be placed on PE3->R5->R6->PE4 as it is allowed to be longer.
Driving the PCE disjointness computation may be done in other ways by for instance setting a metric boundary reflecting an RTD boundary. Other constraints may also be used.The P-Flag allows a simple expression that the disjointness constraint should not make the LSP worst.
Any constraint added to a path disjointness computation may reduce the chance to find suitable paths. The usage of the P-flag, as any other constraint, may prevent to find a disjoint path.
In the example above, if we consider that the router R5 is down, if PE1->PE2 has the P-flag set, there is no room available to place PE3->PE4 (the disjointness constraint cannot be fulfilled).
If PE->PE2 has the P-flag unset, the algorithm may be able to place PE1->PE2 on R1->R2 link leaving a room for PE3->PE4 using the R3->R4 link.
When using P-flag or any additional constraint on top of the disjointness constraint, the user should be aware that there is less chance to fulfill the disjointness constraint.
Multiple LSPs in the same disjoint group may have the P-flag set. In such a case, those LSPs may not be disjoint from each other but will be disjoint from others LSPs in the group that have the P-flag unset.
In the figure above, we still consider the same previous requirements, so PE1->PE2 LSP should be optimized (P-flag set) while PE3->PE4 should be disjoint and may use a longer path.
Regarding PE1->PE2, there are two paths that are satisfying the constraints (ECMP): PE1->R1->R4->R2->PE2 (path 1) and PE1->R1->R3->R4->R2->PE2 (path 2).
An implementation may choose one of the paths or even use both (using both may happen in case Segment Routing TE is used, allowing ECMP).
If the implementation elects only one path, there is a chance that picking up one path may prevent disjointness. In our example, if path 2 is used for PE1->PE2, there is no room left for PE3->PE4 while if path 1 is used, PE3->PE4 can be placed on R3->R4 link.When P-flag is set for an LSP and when ECMPs are available, an implementation MAY select a path that allows disjointness.There may be some cases where the PCE is not able to provide a set of disjoint paths for one or more LSPs in the association.When the T-bit is set (Strict disjointness requested), if disjointness cannot be ensured for one or more LSPs, the PCE SHOULD reply with a PCUpd message containing an empty ERO. In addition to the empty ERO Object, the PCE MAY add the NO-PATH-VECTOR TLV () in the LSP Object.This document adds new bits in the NO-PATH-VECTOR TLV:bit "TBD7": when set, the PCE indicates that it could not find a disjoint path for this LSP.bit "TBD8": when set, the PCE indicates that it does not support the requested disjointness computation.
When the T-bit is unset, the PCE is allowed to reduce the required level of disjointness. The actual level of disjointness computed by the PCE can be reported through the DISJOINTNESS-STATUS-TLV by setting the appropriate flags in the TLV.
While the DISJOINTNESS-CONFIGURATION-TLV defines the expected level of disjointness required by configuration, the DISJOINTNESS-STATUS-TLV defines the actual level of disjointness computed.
There are some cases where the PCE may need to completely relax the disjointness constraint in order to provide a path to all the LSPs that are part of the association.
A mechanism that allows the PCE to fully relax a constraint is considered by the authors as more global to PCEP rather than linked to the disjointness use case. As a consequence, it is considered as out of scope of the document.
All LSPs in a particular disjoint group MUST use the same combination of T,S,N,L flags in the DISJOINTNESS-CONFIGURATION-TLV. If a PCE receives PCRpt messages for LSPs belonging to the same disjoint group but having an inconsistent combination of T,S,N,L flags, the PCE SHOULD NOT try to compute disjointness path and SHOULD reply a PCErr with Error-type 26 (Association Error) and Error-Value 6 (Association information mismatch) to all PCCs involved in the disjoint group.
This document defines one new type for association, which do not add any new
security concerns beyond those discussed in ,
and in itself.
As stated in , much of the information carried in the Disjointness
Association object, as per this document is not extra sensitive. It often reflects information
that can also be derived from the LSP Database, but association provides a much easier grouping of related LSPs and messages. The disjointness association could provides an adversary with the opportunity to eavesdrop on the relationship between the LSPs. Thus securing the PCEP session using Transport Layer
Security (TLS) , as per the recommendations and
best current practices in , is RECOMMENDED.This document defines the following new association type originally
defined in .This document defines the following new PCEP TLVs:
IANA is requested to manage the space of flags carried in the DISJOINTNESS-CONFIGURATION-TLV defined in this document, numbering them from 0 as the least significant bit.
New bit numbers may be allocated in future.IANA is requested to allocate the following bit numbers in the DISJOINTNESS-CONFIGURATION-TLV flag space:three new Objective Functions have been
defined. IANA has made the following allocations from the PCEP
"Objective Function" sub-registry:
This documents defines new bits for the NO-PATH-VECTOR TLV in the
"NO-PATH-VECTOR TLV Flag Field" sub-registry of the "Path Computation
Element Protocol (PCEP) Numbers" registry:
IANA is requested to allocate new Error Types and Error Values within
the " PCEP-ERROR Object Error Types and Values" sub-registry of the
PCEP Numbers registry, as follows:
An operator MUST be allowed to configure the disjointness associations and parameters at PCEP peers
and associate it with the LSPs. describes the PCEP MIB, there are no new MIB Objects
for this document.Mechanisms defined in this document do not imply any new liveness detection
and monitoring requirements in addition to those already listed in
.Mechanisms defined in this document do not imply any new operation
verification requirements in addition to those already listed in
.Mechanisms defined in this document do not imply any new requirements
on other protocols.Mechanisms defined in this document do not have any impact on
network operations in addition to those already listed in
.A special thanks to author of
, this document borrow
some of the text from it. Authors would also like to thank Adrian Farrel for his useful comments.