Internet Engineering Task Force (IETF)                     X. Zhang, Ed.
Request for Comments: 8051                           Huawei Technologies
Category: Informational                                    I. Minei, Ed.
ISSN: 2070-1721                                             Google, Inc.
                                                            January 2017


       Applicability of a Stateful Path Computation Element (PCE)

Abstract

   A stateful Path Computation Element (PCE) maintains information about
   Label Switched Path (LSP) characteristics and resource usage within a
   network in order to provide traffic-engineering calculations for its
   associated Path Computation Clients (PCCs).  This document 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.  PCE Communication
   Protocol (PCEP) extensions required for stateful PCE usage are
   covered in separate documents.

Status of This Memo

   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 a candidate 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
   http://www.rfc-editor.org/info/rfc8051.















Zhang & Minei                 Informational                     [Page 1]


RFC 8051            Applicability for a Stateful PCE        January 2017


Copyright Notice

   Copyright (c) 2017 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
   (http://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.

Table of Contents

   1.  Introduction  . . . . . . . . . . . . . . . . . . . . . . . .   3
   2.  Terminology . . . . . . . . . . . . . . . . . . . . . . . . .   4
   3.  Application Scenarios . . . . . . . . . . . . . . . . . . . .   5
     3.1.  Optimization of LSP Placement . . . . . . . . . . . . . .   5
       3.1.1.  Throughput Maximization and Bin Packing . . . . . . .   6
       3.1.2.  Deadlock  . . . . . . . . . . . . . . . . . . . . . .   7
       3.1.3.  Minimum Perturbation  . . . . . . . . . . . . . . . .   9
       3.1.4.  Predictability  . . . . . . . . . . . . . . . . . . .  10
     3.2.  Auto-Bandwidth Adjustment . . . . . . . . . . . . . . . .  11
     3.3.  Bandwidth Scheduling  . . . . . . . . . . . . . . . . . .  12
     3.4.  Recovery  . . . . . . . . . . . . . . . . . . . . . . . .  12
       3.4.1.  Protection  . . . . . . . . . . . . . . . . . . . . .  13
       3.4.2.  Restoration . . . . . . . . . . . . . . . . . . . . .  14
       3.4.3.  SRLG Diversity  . . . . . . . . . . . . . . . . . . .  15
     3.5.  Maintenance of Virtual Network Topology (VNT) . . . . . .  15
     3.6.  LSP Reoptimization  . . . . . . . . . . . . . . . . . . .  16
     3.7.  Resource Defragmentation  . . . . . . . . . . . . . . . .  17
     3.8.  Point-to-Multipoint Applications  . . . . . . . . . . . .  17
     3.9.  Impairment-Aware Routing and Wavelength Assignment
           (IA-RWA)  . . . . . . . . . . . . . . . . . . . . . . . .  18
   4.  Deployment Considerations . . . . . . . . . . . . . . . . . .  19
     4.1.  Multi-PCE Deployments . . . . . . . . . . . . . . . . . .  19
     4.2.  LSP State Synchronization . . . . . . . . . . . . . . . .  19
     4.3.  PCE Survivability . . . . . . . . . . . . . . . . . . . .  19
   5.  Security Considerations . . . . . . . . . . . . . . . . . . .  20
   6.  References  . . . . . . . . . . . . . . . . . . . . . . . . .  20
     6.1.  Normative References  . . . . . . . . . . . . . . . . . .  20
     6.2.  Informative References  . . . . . . . . . . . . . . . . .  21
   Acknowledgements  . . . . . . . . . . . . . . . . . . . . . . . .  22
   Contributors  . . . . . . . . . . . . . . . . . . . . . . . . . .  22
   Authors' Addresses  . . . . . . . . . . . . . . . . . . . . . . .  24



Zhang & Minei                 Informational                     [Page 2]


RFC 8051            Applicability for a Stateful PCE        January 2017


1.  Introduction

   [RFC4655] defines the architecture for a model based on the Path
   Computation Element (PCE) for the computation of Multiprotocol Label
   Switching (MPLS) and Generalized MPLS (GMPLS) Traffic Engineering
   Label Switched Paths (TE LSPs).  To perform such a constrained
   computation, a PCE stores the network topology (i.e., TE links and
   nodes) and resource information (i.e., TE attributes) in its TE
   Database (TED).  [RFC5440] describes the Path Computation Element
   Protocol (PCEP) for interaction between a Path Computation Client
   (PCC) and a PCE, or between two PCEs, enabling computation of TE
   LSPs.

   As per [RFC4655], a PCE can be either stateful or stateless.  A
   stateful PCE maintains two sets of information for use in path
   computation.  The first is the Traffic Engineering Database (TED),
   which includes the topology and resource state in the network.  This
   information can be obtained by a stateful PCE using the same
   mechanisms as a stateless PCE (see [RFC4655]).  The second is the LSP
   State Database (LSP-DB), in which a PCE stores attributes of all
   active LSPs in the network, such as their paths through the network,
   bandwidth/resource usage, switching types, and LSP constraints.  This
   state information allows the PCE to compute constrained paths while
   considering individual LSPs and their inter-dependency.  However,
   this requires reliable state synchronization mechanisms between the
   PCE and the network, between the PCE and the PCCs, and between
   cooperating PCEs, with potentially significant control-plane overhead
   and maintenance of a large amount of state data, as explained in
   [RFC4655].

   This document describes how a stateful PCE can be used to solve
   various problems for MPLS-TE and GMPLS networks and the benefits it
   brings to such deployments.  Note that alternative solutions relying
   on stateless PCEs may also be possible for some of these use cases
   and will be mentioned for completeness where appropriate.
















Zhang & Minei                 Informational                     [Page 3]


RFC 8051            Applicability for a Stateful PCE        January 2017


2.  Terminology

   This document uses the following terms defined in [RFC5440]: PCC,
   PCE, and PCEP peer.

   This document defines the following terms:

   Stateful PCE:  a PCE that has access to not only the network state,
      but also to the set of active paths and their reserved resources
      for its computations.  A stateful PCE might also retain
      information regarding LSPs under construction in order to reduce
      churn and resource contention.  The additional state allows the
      PCE to compute constrained paths while considering individual LSPs
      and their interactions.  Note that this requires reliable state
      synchronization mechanisms between the PCE and the network, PCE
      and PCC, and between cooperating PCEs.

   Passive Stateful PCE:  a PCE that uses LSP state information learned
      from PCCs to optimize path computations.  It does not actively
      update LSP state.  A PCC maintains synchronization with the PCE.

   Active Stateful PCE:  a PCE that may issue recommendations to the
      network.  For example, an Active Stateful PCE may use the
      Delegation mechanism to update LSP parameters in those PCCs that
      delegate control over their LSPs to the PCE.

   Delegation:  an operation to grant a PCE temporary rights to modify a
      subset of LSP parameters on one or more LSPs of a PCC.  LSPs are
      delegated from a PCC to a PCE and are referred to as "delegated"
      LSPs.  The PCC that owns the PCE state for the LSP has the right
      to delegate it.  An LSP is owned by a single PCC at any given
      point in time.  For intra-domain LSPs, this PCC should be the LSP
      head end.

   LSP State Database:  information about all LSPs and their attributes.

   PCE Initiation:  assuming LSP delegation granted by default, a PCE
      can issue recommendations to the network.

   Minimum Cut Set:  the minimum set of links for a specific source
      destination pair that, when removed from the network, results in a
      specific source being completely isolated from a specific
      destination.  The summed capacity of these links is equivalent to
      the maximum capacity from the source to the destination by the
      max-flow min-cut theorem.






Zhang & Minei                 Informational                     [Page 4]


RFC 8051            Applicability for a Stateful PCE        January 2017


3.  Application Scenarios

   In the following sections, several use cases are described,
   showcasing scenarios that benefit from the deployment of a stateful
   PCE.

3.1.  Optimization of LSP Placement

   The following use cases demonstrate a need for visibility into global
   LSP states in PCE path computations, and for a PCE control of
   sequence and timing in altering LSP path characteristics within and
   across PCEP sessions.  Reference topologies for the use cases
   described later in this section are shown in Figures 1 and 2.

   Some of the use cases below are focused on MPLS-TE deployments but
   may also apply to GMPLS.  Unless otherwise cited, use cases assume
   that all LSPs listed exist at the same LSP priority.

   The main benefit in the cases below comes from moving away from an
   asynchronous PCC-driven mode of operation to a model that allows for
   central control over LSP computations and maintenance, and focuses
   specifically on the active stateful PCE model of operation.

          +-----+
          |  A  |
          +-----+
                 \
                  +-----+                      +-----+
                  |  C  |----------------------|  E  |
                  +-----+                      +-----+
                 /        \      +-----+      /
          +-----+          +-----|  D  |-----+
          |  B  |                +-----+
          +-----+

                      Figure 1: Reference Topology 1

               +-----+        +-----+        +-----+
               |  A  |        |  B  |        |  C  |
               +--+--+        +--+--+        +--+--+
                  |              |              |
                  |              |              |
               +--+--+        +--+--+        +--+--+
               |  E  +--------+  F  +--------+  G  |
               +-----+        +-----+        +-----+

                      Figure 2: Reference Topology 2




Zhang & Minei                 Informational                     [Page 5]


RFC 8051            Applicability for a Stateful PCE        January 2017


3.1.1.  Throughput Maximization and Bin Packing

   Because LSP attribute changes in [RFC5440] are driven by Path
   Computation Request (PCReq) messages under control of a PCC's local
   timers, the sequence of resource reservation arrivals occurring in
   the network will be randomized.  This, coupled with a lack of global
   LSP state visibility on the part of a stateless PCE, may result in
   suboptimal throughput in a given network topology, as will be shown
   in the example below.

   Reference Topology 2 in Figure 2 and Tables 1 and 2 show an example
   in which throughput is at 50% of optimal as a result of the lack of
   visibility and synchronized control across PCCs.  In this scenario,
   the decision must be made as to whether to route any portion of the
   E-G demand, as any demand routed for this source and destination will
   decrease system throughput.

                       +------+--------+----------+
                       | Link | Metric | Capacity |
                       +------+--------+----------+
                       | A-E  |   1    |    10    |
                       | B-F  |   1    |    10    |
                       | C-G  |   1    |    10    |
                       | E-F  |   1    |    10    |
                       | F-G  |   1    |    10    |
                       +------+--------+----------+

             Table 1: Link Parameters for Throughput Use Case

          +------+-----+-----+-----+--------+----------+-------+
          | Time | LSP | Src | Dst | Demand | Routable |  Path |
          +------+-----+-----+-----+--------+----------+-------+
          |  1   |  1  |  E  |  G  |   10   |   Yes    | E-F-G |
          |  2   |  2  |  A  |  B  |   10   |    No    |  ---  |
          |  3   |  1  |  F  |  C  |   10   |    No    |  ---  |
          +------+-----+-----+-----+--------+----------+-------+

              Table 2: Throughput Use Case Demand Time Series

   In many cases, throughput maximization becomes a bin-packing problem.
   While bin packing itself is an NP-hard problem, a number of common
   heuristics that run in polynomial time can provide significant
   improvements in throughput over random reservation event
   distribution, especially when traversing links that are members of
   the minimum cut set for a large subset of source destination pairs.






Zhang & Minei                 Informational                     [Page 6]


RFC 8051            Applicability for a Stateful PCE        January 2017


   Tables 3 and 4 show a simple use case using Reference Topology 1 in
   Figure 1, where LSP state visibility and control of reservation order
   across PCCs would result in significant improvement in total
   throughput.

                       +------+--------+----------+
                       | Link | Metric | Capacity |
                       +------+--------+----------+
                       | A-C  |   1    |    10    |
                       | B-C  |   1    |    10    |
                       | C-E  |   10   |    5     |
                       | C-D  |   1    |    10    |
                       | D-E  |   1    |    10    |
                       +------+--------+----------+

             Table 3: Link Parameters for Bin-Packing Use Case

         +------+-----+-----+-----+--------+----------+---------+
         | Time | LSP | Src | Dst | Demand | Routable |   Path  |
         +------+-----+-----+-----+--------+----------+---------+
         |  1   |  1  |  A  |  E  |   5    |   Yes    | A-C-D-E |
         |  2   |  2  |  B  |  E  |   10   |    No    |   ---   |
         +------+-----+-----+-----+--------+----------+---------+

             Table 4: Bin-Packing Use Case Demand Time Series

3.1.2.  Deadlock

   This section discusses the use case of cross-LSP impact under
   degraded operation.  Most existing RSVP-TE implementations will not
   tear down established LSPs in the event of the failure of the
   bandwidth increase procedure detailed in [RFC3209].  This behavior is
   directly implied to be correct in [RFC3209] and is often desirable
   from an operator's perspective, because either a) the destination
   prefixes are not reachable via any means other than MPLS or b) this
   would result in significant packet loss as demand is shifted to other
   LSPs in the overlay mesh.

   In addition, there are currently few implementations offering dynamic
   ingress admission control (policing of the traffic volume mapped onto
   an LSP) at the Label Edge Router (LER).  Having ingress admission
   control on a per-LSP basis is not necessarily desirable from an
   operational perspective, as a) one must over-provision tunnels
   significantly in order to avoid deleterious effects resulting from
   stacked transport and flow control systems (for example, for tunnels
   that are dynamically resized based on current traffic) and b) there
   is currently no efficient commonly available northbound interface for
   dynamic configuration of per-LSP ingress admission control.



Zhang & Minei                 Informational                     [Page 7]


RFC 8051            Applicability for a Stateful PCE        January 2017


   Lack of ingress admission control coupled with the behavior in
   [RFC3209] may result in LSPs operating out of profile for significant
   periods of time.  It is reasonable to expect that these out-of-
   profile LSPs will be operating in a degraded state and experience
   traffic loss.  Moreover, because those LSPs end up sharing common
   network interfaces with other LPSs operating within their bandwidth
   reservations, they will impact the operation of the in-profile LSPs,
   even when there is unused network capacity elsewhere in the network.
   Furthermore, this behavior will cause information loss in the TED
   with regards to the actual available bandwidth on the links used by
   the out-of-profile LSPs, as the reservations on the links no longer
   reflect the capacity used.

   Reference Topology 1 in Figure 1 and Tables 5 and 6 show a use case
   that demonstrates this behavior.  Two LSPs, LSP 1 and LSP 2, are
   signaled with demand 2 and routed along paths A-C-D-E and B-C-D-E,
   respectively.  At a later time, the demand of LSP 1 increases to 20.
   Under such a demand, the LSP cannot be resignaled.  However, the
   existing LSP will not be torn down.  In the absence of ingress
   policing, traffic on LSP 1 will cause degradation for traffic of LSP
   2 (due to oversubscription on the links C-D and D-E), as well as
   information loss in the TED with regard to the actual network state.

   The problem could be easily ameliorated by global visibility of the
   LSP state coupled with PCC-external demand measurements and placement
   of two LSPs on disjoint links.  Note that while the demand of 20 for
   LSP 1 could never be satisfied in the given topology, isolation from
   the ill-effects of the (unsatisfiable) increased demand could be
   achieved.

                       +------+--------+----------+
                       | Link | Metric | Capacity |
                       +------+--------+----------+
                       | A-C  |   1    |    10    |
                       | B-C  |   1    |    10    |
                       | C-E  |   10   |    5     |
                       | C-D  |   1    |    10    |
                       | D-E  |   1    |    10    |
                       +------+--------+----------+

       Table 5: Link Parameters for the 'Degraded Operation' Example










Zhang & Minei                 Informational                     [Page 8]


RFC 8051            Applicability for a Stateful PCE        January 2017


         +------+-----+-----+-----+--------+----------+---------+
         | Time | LSP | Src | Dst | Demand | Routable |   Path  |
         +------+-----+-----+-----+--------+----------+---------+
         |  1   |  1  |  A  |  E  |   2    |   Yes    | A-C-D-E |
         |  2   |  2  |  B  |  E  |   2    |   Yes    | B-C-D-E |
         |  3   |  1  |  A  |  E  |   20   |    No    |   ---   |
         +------+-----+-----+-----+--------+----------+---------+

             Table 6: 'Degraded Operation' Demand Time Series

3.1.3.  Minimum Perturbation

   As a result of both the lack of visibility into the global LSP state
   and the lack of control over event ordering across PCE sessions,
   unnecessary perturbations may be introduced into the network by a
   stateless PCE.  Tables 7 and 8 show an example of an unnecessary
   network perturbation using Reference Topology 1 in Figure 1.  In this
   case, an unimportant (high LSP priority value) LSP (LSP1) is first
   set up along the shortest path.  At time 2, which is assumed to be
   relatively close to time 1, a second more important (lower LSP-
   priority value) LSP (LSP2) is established, preempting LSP1
   potentially causing traffic loss.  LSP1 is then reestablished on the
   longer A-C-E path.

                       +------+--------+----------+
                       | Link | Metric | Capacity |
                       +------+--------+----------+
                       | A-C  |   1    |    10    |
                       | B-C  |   1    |    10    |
                       | C-E  |   10   |    10    |
                       | C-D  |   1    |    10    |
                       | D-E  |   1    |    10    |
                       +------+--------+----------+

      Table 7: Link Parameters for the 'Minimum-Perturbation' Example

    +------+-----+-----+-----+--------+----------+----------+---------+
    | Time | LSP | Src | Dst | Demand | LSP Prio | Routable |   Path  |
    +------+-----+-----+-----+--------+----------+----------+---------+
    |  1   |  1  |  A  |  E  |   7    |    7     |   Yes    | A-C-D-E |
    |  2   |  2  |  B  |  E  |   7    |    0     |   Yes    | B-C-D-E |
    |  3   |  1  |  A  |  E  |   7    |    7     |   Yes    |  A-C-E  |
    +------+-----+-----+-----+--------+----------+----------+---------+

        Table 8: 'Minimum-Perturbation' LSP and Demand Time Series






Zhang & Minei                 Informational                     [Page 9]


RFC 8051            Applicability for a Stateful PCE        January 2017


   A stateful PCE can help in this scenario by computing both routes at
   the same time.  The advantages of using a stateful PCE over
   exploiting a stateless PCE via Global Concurrent Optimization (GCO)
   are threefold.  First is the ability to accommodate concurrent path
   computation from different PCCs.  Second is the reduction of control-
   plane overhead since the stateful PCE has the route information of
   the affected LSPs.  Thirdly, the stateful PCE can use the LSP-DB to
   further optimize the placement of LSPs.  This will ensure placement
   of the more important LSP along the shortest path, avoiding the setup
   and subsequent preemption of the lower priority LSP.  Similarly, when
   a new higher priority LSP that requires preemption of an existing
   lower priority LSP(s), a stateful PCE can determine the minimum
   number of lower priority LSPs to reroute using the Make-Before-Break
   (MBB) mechanism without disrupting any service and then set up the
   higher priority LSP.

3.1.4.  Predictability

   Randomization of reservation events caused by lack of control over
   event ordering across PCE sessions results in poor predictability in
   LSP routing.  An offline system applying a consistent optimization
   method will produce predictable results to within either the boundary
   of forecast error (when reservations are over-provisioned by
   reasonable margins) or to the variability of the signal and the
   forecast error (when applying some hysteresis in order to minimize
   churn).  Predictable results are valuable for being able to simulate
   the network and reliably test it under various scenarios, especially
   under various failure modes and planned maintenances when predictable
   path characteristics are desired under contention for network
   resources.

   Reference Topology 1 and Tables 9, 10, and 11 show the impact of
   event ordering and predictability of LSP routing.

                       +------+--------+----------+
                       | Link | Metric | Capacity |
                       +------+--------+----------+
                       | A-C  |   1    |    10    |
                       | B-C  |   1    |    10    |
                       | C-E  |   1    |    10    |
                       | C-D  |   1    |    10    |
                       | D-E  |   1    |    10    |
                       +------+--------+----------+

         Table 9: Link Parameters for the 'Predictability' Example






Zhang & Minei                 Informational                    [Page 10]


RFC 8051            Applicability for a Stateful PCE        January 2017


         +------+-----+-----+-----+--------+----------+---------+
         | Time | LSP | Src | Dst | Demand | Routable |   Path  |
         +------+-----+-----+-----+--------+----------+---------+
         |  1   |  1  |  A  |  E  |   7    |   Yes    |  A-C-E  |
         |  2   |  2  |  B  |  E  |   7    |   Yes    | B-C-D-E |
         +------+-----+-----+-----+--------+----------+---------+

          Table 10: 'Predictability' LSP and Demand Time Series 1

         +------+-----+-----+-----+--------+----------+---------+
         | Time | LSP | Src | Dst | Demand | Routable |   Path  |
         +------+-----+-----+-----+--------+----------+---------+
         |  1   |  2  |  B  |  E  |   7    |   Yes    |  B-C-E  |
         |  2   |  1  |  A  |  E  |   7    |   Yes    | A-C-D-E |
         +------+-----+-----+-----+--------+----------+---------+

          Table 11: 'Predictability' LSP and Demand Time Series 2

   As can be shown in the example, both LSPs are routed in both cases,
   but along very different paths.  This would be a challenge if
   reliable simulation of the network is attempted.  An active stateful
   PCE can solve this through control over LSP ordering.  Based on
   triggers such as a failure or an optimization trigger, the PCE can
   order the computations and path setup in a deterministic way.

3.2.  Auto-Bandwidth Adjustment

   The bandwidth requirements of LSPs often change over time, requiring
   LSP resizing.  In most implementations available today, the head-end
   node performs this function by monitoring the actual bandwidth usage,
   triggering a recomputation and resignaling when a threshold is
   reached.  This operation is referred to as "auto-bandwidth
   adjustment".  The head-end node either recomputes the path locally,
   or it requests a recomputation from a PCE by sending a PCReq message.
   In the latter case, the PCE computes a new path and provides the new
   route suggestion.  Upon receiving the reply from the PCE, the PCC
   resignals the LSP in Shared-Explicit (SE) mode along the newly
   computed path.  With a stateless PCE, the head-end node needs to
   provide the currently used bandwidth and the route information via
   path computation request messages.  Note that in this scenario, the
   head-end node is the one that drives the LSP resizing based on local
   information, and that the difference between using a stateless and a
   passive stateful PCE is in the level of optimization of the LSP
   placement as discussed in the previous section.

   A more interesting smart bandwidth adjustment case is one where the
   LSP resizing decision is done by an external entity with access to
   additional information such as historical trending data, application-



Zhang & Minei                 Informational                    [Page 11]


RFC 8051            Applicability for a Stateful PCE        January 2017


   specific information about expected demands or policy information, as
   well as knowledge of the actual desired flow volumes.  In this case,
   an active stateful PCE provides an advantage in both the computation
   with knowledge of all LSPs in the domain and in the ability to
   trigger bandwidth modification of the LSP.

3.3.  Bandwidth Scheduling

   Bandwidth scheduling allows network operators to reserve resources in
   advance according to the agreements with their customers and allows
   them to transmit data with a specified starting time and duration,
   for example, for a scheduled bulk data replication between data
   centers.

   Traditionally, this can be supported by Network Management System
   (NMS) operation through path pre-establishment and activation on the
   agreed starting time.  However, this does not provide efficient
   network usage since the established paths exclude the possibility of
   being used by other services even when they are not used for
   undertaking any service.  It can also be accomplished through GMPLS
   protocol extensions by carrying the related request information
   (e.g., starting time and duration) across the network.  Nevertheless,
   this method inevitably increases the complexity of the signaling and
   routing process.

   A passive stateful PCE can support this application with better
   efficiency since it can alleviate the burden of processing on network
   elements.  This requires the PCE to maintain the scheduled LSPs and
   their associated resource usage, as well as the ability of head-ends
   to trigger signaling for LSP setup/deletion at the correct time.
   This approach requires coarse time synchronization between PCEs and
   PCCs.  With PCE initiation capability, a PCE can trigger the setup
   and deletion of scheduled requests in a centralized manner, without
   modification of existing head-end behaviors, by notifying the PCCs to
   set up or tear down the paths.

3.4.  Recovery

   The recovery use cases discussed in the following sections show how
   leveraging a stateful PCE can simplify the computation of recovery
   path(s).  In particular, two characteristics of a stateful PCE are
   used: 1) using information stored in the LSP-DB for determining
   shared protection resources and 2) performing computations with
   knowledge of all LSPs in a domain.







Zhang & Minei                 Informational                    [Page 12]


RFC 8051            Applicability for a Stateful PCE        January 2017


3.4.1.  Protection

   If a PCC can specify in a request whether the computation is for a
   working path or for protection and a PCC can report the resource as a
   working or protection path, then the following text applies.  A PCC
   can send multiple requests to the PCE, asking for two LSPs, and use
   them as working and backup paths separately.  Either way, the
   resources bound to backup paths can be shared by different LSPs to
   improve the overall network efficiency, such as m:n protection or
   pre-configured shared mesh recovery techniques as specified in
   [RFC4427].  If resource sharing is supported for LSP protection, the
   information relating to existing LSPs is required to avoid allocation
   of shared protection resources to two LSPs that might fail together
   and cause protection contention issues.  A stateless PCE can
   accommodate this use case by having the PCC pass this information as
   a constraint in the path computation request.  A passive stateful PCE
   can more easily accommodate this need using the information stored in
   its LSP-DB.  Furthermore, an active stateful PCE can help with
   (re)optimization of protection resource sharing as well as LSP
   maintenance operation with less impact on protection resources.

                 +----+
                 |PCE |
                 +----+

            +------+          +------+          +------+
            |  A   +----------+  B   +----------+  C   |
            +--+---+          +---+--+          +---+--+
               |                  |                 |
               |        +---------+                 |
               |        |                           |
               |     +--+---+          +------+     |
               +-----+  E   +----------+  D   +-----+
                     +------+          +------+

                      Figure 3: Reference Topology 3

   For example, in the network depicted in Figure 3, suppose there
   exists LSP1 with working path LSP1_working following A->E and with
   backup path LSP1_backup following A->B->E.  A request arrives asking
   for a working and backup path pair to be computed for LSP2 from B to
   E.  If the PCE decides LSP2_working follows B->A->E, then the backup
   path LSP2_backup should not share the same protection resource with
   LSP1 since LSP2 shares part of its resource (specifically A->E) with
   LSP1 (i.e., these two LSPs are in the same shared risk group).  There
   is no such constraint if B->C->D->E is chosen for LSP2_working.





Zhang & Minei                 Informational                    [Page 13]


RFC 8051            Applicability for a Stateful PCE        January 2017


   If a stateless PCE is used, the head node B needs to be aware of the
   existence of LSPs that share the route of LSP2_working and of the
   details of their protection resources.  B must pass this information
   to the PCE as a constraint so as to request a path with diversity.
   Alternatively, a stateless PCE may be able to compute paths
   diversified by SRLG (Shared Risk Link Group) if TED is extended so
   that it includes the SRLG information that is protected by a given
   backup resource, but at the expense of a high complexity in routing.
   On the other hand, a stateful PCE can get the LSPs information by
   itself given the LSP identifier(s) and can then find SRLG-diversified
   protection paths for both LSPs.  This is made possible by comparing
   the LSP resource usage exploiting the LSP-DB accessible by the
   stateful PCE.

3.4.2.  Restoration

   In case of a link failure, such as a fiber cut, multiple LSPs may
   fail at the same time.  Thus, the source nodes of the affected LSPs
   will be informed of the failure by the nodes detecting the failure.
   These source nodes will send requests to a PCE for rerouting.  In
   order to reuse the resource taken by an existing LSP, the source node
   can send a PCReq message that includes the Exclude Route Object (XRO)
   with Fail (F) bit set together with the Record Route Object (RRO)
   that contains the current route information, as specified in
   [RFC5521].

   If a stateless PCE is used, it might respond to the rerouting
   requests separately if the requests arrive at different times.  Thus,
   it might result in suboptimal resource usage.  Even worse, it might
   unnecessarily block some of the rerouting requests due to
   insufficient resources for rerouting messages that arrive later.  If
   a passive stateful PCE is used to fulfill this task, the procedure
   can be simplified.  The PCCs reporting the failures can include LSP
   identifiers instead of detailed information, and the PCE can find
   relevant LSP information by inspecting the LSP-DB.  Moreover, the PCE
   can recompute the affected LSPs concurrently while reusing part of
   the existing LSP's resources when it is informed of the failed link
   identifier provided by the first request.  This is made possible
   because the passive stateful PCE can check what other LSPs are
   affected by the failed link and their route information by inspecting
   its LSP-DB.  As a result, a better performance can be achieved, such
   as better resource usage or minimal probability of blocking upcoming
   new rerouting requests sent as a result of the link failure.

   If the target is to avoid resource contention within the time window
   of a high number of LSP rerouting requests, a stateful PCE can retain
   the under-construction LSP resource usage information for a given
   time and exclude it from being used for a forthcoming LSP's request.



Zhang & Minei                 Informational                    [Page 14]


RFC 8051            Applicability for a Stateful PCE        January 2017


   In this way, it can ensure that the resource will not be double-
   booked; thus, the issue of resource contention and computation crank-
   backs can be alleviated.

3.4.3.  SRLG Diversity

   An alternative way to achieve efficient resilience is to maintain
   SRLG disjointness between LSPs, irrespective of whether or not these
   LSPs share the source and destination nodes.  This can be achieved at
   provisioning time, if the routes of all the LSPs are requested
   together, using a synchronized computation of the different LSPs with
   SRLG disjointness constraint.  If the LSPs need to be provisioned at
   different times, the PCC can specify, as constraints to the path
   computation, a set of SRLGs using the Exclude Route Object [RFC5521].
   However, for the latter to be effective, the entity that requests the
   route to the PCE needs to maintain updated SRLG information regarding
   all of the LSPs to which it must maintain the disjointness.  A
   stateless PCE can compute an SRLG-disjoint path by inspecting the TED
   and precluding the links with the same SRLG values specified in the
   PCReq message sent by a PCC.

   A passive stateful PCE maintains the updated SRLG information of the
   established LSPs in a centralized manner.  Therefore, the PCC can
   specify, as constraints to the path computation, the SRLG
   disjointness of a set of already established LSPs by only providing
   the LSP identifiers.  Similarly, a passive stateful PCE can also
   accommodate disjointness using other constraints, such as link, node,
   or path segment.

3.5.  Maintenance of Virtual Network Topology (VNT)

   In Multi-Layer Networks (MLN), a Virtual Network Topology (VNT)
   [RFC5212] consists of a set of one or more TE LSPs in the lower
   layer, which provides TE links to the upper layer.  In [RFC5623], the
   PCE-based architecture is proposed to support path computation in MLN
   networks in order to achieve inter-layer TE.

   The establishment/teardown of a TE link in VNT needs to take into
   consideration the state of existing LSPs and/or new LSP request(s) in
   the higher layer.  Hence, when a stateless PCE cannot find the route
   for a request based on the upper-layer topology information, it does
   not have enough information to decide whether or not to set up or
   remove a TE link, which then can result in non-optimal usage of a
   resource.  On the other hand, a passive stateful PCE can make a
   better decision of when and how to modify the VNT either to
   accommodate new LSP requests or to reoptimize resource usage across
   layers irrespective of the PCE models as described in [RFC5623].
   Furthermore, given the active capability, the stateful PCE can issue



Zhang & Minei                 Informational                    [Page 15]


RFC 8051            Applicability for a Stateful PCE        January 2017


   VNT modification suggestions in order to accommodate path setup
   requests or reoptimize resource usage across layers.

3.6.  LSP Reoptimization

   In order to make efficient usage of network resources, it is
   sometimes desirable to reoptimize one or more LSPs dynamically.  In
   the case of a stateless PCE, in order to optimize network resource
   usage dynamically through online planning, a PCC must send a request
   to the PCE together with detailed path/bandwidth information of the
   LSPs that need to be concurrently optimized.  This means that the PCC
   must be able to determine when and which LSPs should be optimized.
   In the case of a passive stateful PCE, given the LSP state
   information in the LSP database, the process of dynamic optimization
   of network resources can be simplified without requiring the PCC to
   supply detailed LSP state information.  Moreover, an active stateful
   PCE can even make the process automated by triggering the request.
   Because a stateful PCE can maintain information for all LSPs that are
   in the process of being set up and it may have the ability to control
   timing and sequence of LSP setup/deletion, the optimization
   procedures can be performed more intelligently and effectively.  A
   stateful PCE can also determine which LSP should be reoptimized based
   on network events.  For example, when an LSP is torn down, its
   resources are freed.  This can trigger the stateful PCE to
   automatically determine which LSP should be reoptimized so that the
   recently freed resources may be allocated to it.

   A special case of LSP reoptimization is GCO [RFC5557].  Global
   control of the LSP operation sequence in [RFC5557] is predicated on
   the use of what is effectively a stateful (or semi-stateful) NMS.
   The NMS can be either not local to the network nodes, in which case
   another northbound interface is required for LSP attribute changes,
   or local/collocated, in which case there are significant issues with
   efficiency in resource usage.  A stateful PCE adds a few features
   that:

   o  Roll the NMS visibility into the PCE and remove the requirement
      for an additional northbound interface.

   o  Allow the PCE to determine when reoptimization is needed, with
      which level (GCO or a more incremental optimization).

   o  Allow the PCE to determine which LSPs should be reoptimized.

   o  Allow a PCE to control the sequence of events across multiple
      PCCs, allowing for bulk (and truly global) optimization, LSP
      shuffling, etc.




Zhang & Minei                 Informational                    [Page 16]


RFC 8051            Applicability for a Stateful PCE        January 2017


3.7.  Resource Defragmentation

   If LSPs are dynamically allocated and released over time, the
   resource becomes fragmented.  In networks with link bundle, the
   overall available resource on a (bundle) link might be sufficient for
   a new LSP request, but if the available resource is not continuous,
   the request is rejected.  Stateful PCEs can be used to perform the
   defragmentation procedure, because global visibility of LSPs in the
   network is required to accurately assess resources on the LSPs and to
   perform defragmentation while ensuring a minimal disruption of the
   network.  This use case cannot be accommodated by a stateless PCE
   because it does not possess the detailed information of existing LSPs
   in the network.

   Another case of particular interest is the optical spectrum
   defragmentation in flexible-grid networks.  In flexible-grid networks
   [RFC7698], LSPs with different optical spectrum sizes (such as
   12.5GHz, 25GHz, etc.) can coexist so as to accommodate the services
   with different bandwidth requests.  Therefore, even if the overall
   spectrum size can meet the service request, it may not be usable if
   the available spectrum resource is not contiguous, but rather
   fragmented into smaller pieces.  Thus, with the help of existing LSP
   state information, a stateful PCE can make the resource grouped
   together to be usable.  Moreover, a stateful PCE can proactively
   choose routes for upcoming path requests to reduce the chance of
   spectrum fragmentation.

3.8.  Point-to-Multipoint Applications

   PCE has been identified as an appropriate technology for the
   determination of the paths of Point-to-Multipoint (P2MP) TE LSPs
   [RFC5671].  The application scenarios and use cases described in
   Sections 3.1, 3.4, and 3.6 are also applicable to P2MP TE LSPs.

   In addition to these, the stateful nature of a PCE simplifies the
   information conveyed in PCEP messages since it is possible to refer
   to the LSPs via an identifier.  For P2MP, this is an added advantage
   where the size of the PCEP message is much larger.  In case of
   stateless PCEs, modification of a P2MP tree requires encoding of all
   leaves along with the paths in a PCReq message.  But by using a
   stateful PCE with P2MP capability, the PCEP message can be used to
   convey only the modifications (the other information can be retrieved
   from the identifier via the LSP-DB).








Zhang & Minei                 Informational                    [Page 17]


RFC 8051            Applicability for a Stateful PCE        January 2017


3.9.  Impairment-Aware Routing and Wavelength Assignment (IA-RWA)

   In Wavelength Switched Optical Networks (WSONs) [RFC6163], a
   wavelength-switched LSP traverses one or more fiber links.  The bit
   rates of the client signals carried by the wavelength LSPs may be the
   same or different.  Hence, a fiber link may transmit a number of
   wavelength LSPs with equal or mixed bit-rate signals.  For example, a
   fiber link may multiplex the wavelengths with only 10 Gbit/s signals,
   mixed 10 Gbit/s and 40 Gbit/s signals, or mixed 40 Gbit/s and 100
   Gbit/s signals.

   IA-RWA in WSONs refers to the process (i.e., lightpath computation)
   that takes into account the optical layer/transmission imperfections
   as additional (i.e., physical layer) constraints.  To be more
   specific, linear and non-linear effects associated with the optical
   network elements should be incorporated into the route and wavelength
   assignment procedure.  For example, the physical imperfection can
   result in the interference of two adjacent lightpaths.  Thus, a guard
   band should be reserved between them to alleviate these effects.  The
   width of the guard band between two adjacent wavelengths depends on
   their characteristics, such as modulation formats and bit rates.  Two
   adjacent wavelengths with different characteristics (e.g., different
   bit rates) may need a wider guard band and those with the same
   characteristics may need a narrower guard band.  For example, 50 GHz
   spacing may be acceptable for two adjacent wavelengths with 40 G
   signals.  But for two adjacent wavelengths with different bit rates
   (e.g., 10 G and 40 G), a larger spacing such as 300 GHz may be
   needed.  Hence, the characteristics (states) of the existing
   wavelength LSPs should be considered for a new RWA request in WSON.

   In summary, when stateful PCEs are used to perform the IA-RWA
   procedure, they need to know the characteristics of the existing
   wavelength LSPs.  The impairment information relating to existing and
   to-be-established LSPs can be obtained by nodes in WSON networks via
   external configuration or other means such as monitoring or
   estimation based on a vendor-specific impair model.  However, WSON-
   related routing protocols, i.e., [RFC7688] and [RFC7580], only
   advertise limited information (i.e., availability) of the existing
   wavelengths, without defining the supported client bit rates.  It
   will incur a substantial amount of control-plane overhead if routing
   protocols are extended to support dissemination of the new
   information relevant for the IA-RWA process.  In this scenario,
   stateful PCE(s) would be a more appropriate mechanism to solve this
   problem.  Stateful PCE(s) can exploit impairment information of LSPs
   stored in LSP-DB to provide accurate RWA calculation.






Zhang & Minei                 Informational                    [Page 18]


RFC 8051            Applicability for a Stateful PCE        January 2017


4.  Deployment Considerations

   This section discusses general issues with stateful PCE deployments
   and identifies areas where additional protocol extensions and
   procedures are needed to address them.  Definitions of protocol
   mechanisms are beyond the scope of this document.

4.1.  Multi-PCE Deployments

   Stateless and stateful PCEs can coexist in the same network and be in
   charge of path computation of different types.  To solve the problem
   of distinguishing between the two types of PCEs, either discovery or
   configuration may be used.

   Multiple stateful PCEs can coexist in the same network.  These PCEs
   may provide redundancy for load sharing, resilience, or partitioning
   of computation features.  Regardless of the reason for multiple PCEs,
   an LSP is only delegated to one of the PCEs at any given point in
   time.  However, an LSP can be redelegated between PCEs, for example,
   when a PCE fails.  [RFC7399] discusses various approaches for
   synchronizing state among the PCEs when multiple PCEs are used for
   load sharing or backup and compute LSPs for the same network.

4.2.  LSP State Synchronization

   The LSP-DB is populated using information received from the PCC.
   Because the accuracy of the computations depends on the accuracy of
   the databases used and because the updates must reach the PCE from
   the network, it is worth noting that the PCE view lags behind the
   true state of the network.  Thus, the use of stateful PCE reduces but
   cannot eliminate the possibility of crankbacks, nor can it guarantee
   optimal computations all the time.  [RFC7399] discusses these
   limitations and potential ways to alleviate them.

   In case of multiple PCEs with different capabilities coexisting in
   the same network, such as a passive stateful PCE and an active
   stateful PCE, it is useful to refer to an LSP, be it delegated or
   not, by a unique identifier instead of providing detailed information
   (e.g., route, bandwidth) associated with it, when these PCEs
   cooperate on path computation, such as for load sharing.

4.3.  PCE Survivability

   For a stateful PCE, an important issue is to get the LSP state
   information resynchronized after a restart.  LSP state
   synchronization procedures can be applied equally to a network node
   or another PCE, allowing multiple ways to reacquire the LSP database
   on a restart.  Because synchronization may also be skipped, if a PCE



Zhang & Minei                 Informational                    [Page 19]


RFC 8051            Applicability for a Stateful PCE        January 2017


   implementation has the means to retrieve its database in a different
   way (for example, from a backup copy stored locally), the state can
   be restored without further overhead in the network.  A hybrid
   approach where the bulk of the state is recovered locally, and a
   small amount of state is reacquired from the network, is also
   possible.  Note that locally recovering the state would still require
   some degree of resynchronization to ensure that the recovered state
   is indeed up-to-date.  Depending on the resynchronization mechanism
   used, there may be an additional load on the PCE, and there may be a
   delay in reaching the synchronized state, which may negatively affect
   survivability.  Different resynchronization methods are suited for
   different deployments and objectives.

5.  Security Considerations

   This document 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.  No new
   protocol extensions to PCEP are defined in this document.

   The PCEP extensions in support of the stateful PCE and the delegation
   of path control ability can result in more information and control
   being available for a hypothetical adversary and a number of
   additional attack surfaces that must be protected.  This includes,
   but is not limited to, the authentication and encryption of PCEP
   sessions, snooping of the state of the LSPs active in the network,
   etc.  Therefore, documents in which the PCEP protocol extensions are
   defined need to consider the issues and risks associated with a
   stateful PCE.

6.  References

6.1.  Normative References

   [RFC4655]  Farrel, A., Vasseur, J., and J. Ash, "A Path Computation
              Element (PCE)-Based Architecture", RFC 4655,
              DOI 10.17487/RFC4655, August 2006,
              <http://www.rfc-editor.org/info/rfc4655>.

   [RFC5440]  Vasseur, JP., Ed. and JL. Le Roux, Ed., "Path Computation
              Element (PCE) Communication Protocol (PCEP)", RFC 5440,
              DOI 10.17487/RFC5440, March 2009,
              <http://www.rfc-editor.org/info/rfc5440>.

   [RFC7399]  Farrel, A. and D. King, "Unanswered Questions in the Path
              Computation Element Architecture", RFC 7399,
              DOI 10.17487/RFC7399, October 2014,
              <http://www.rfc-editor.org/info/rfc7399>.



Zhang & Minei                 Informational                    [Page 20]


RFC 8051            Applicability for a Stateful PCE        January 2017


6.2.  Informative References

   [RFC3209]  Awduche, D., Berger, L., Gan, D., Li, T., Srinivasan, V.,
              and G. Swallow, "RSVP-TE: Extensions to RSVP for LSP
              Tunnels", RFC 3209, DOI 10.17487/RFC3209, December 2001,
              <http://www.rfc-editor.org/info/rfc3209>.

   [RFC4427]  Mannie, E., Ed. and D. Papadimitriou, Ed., "Recovery
              (Protection and Restoration) Terminology for Generalized
              Multi-Protocol Label Switching (GMPLS)", RFC 4427,
              DOI 10.17487/RFC4427, March 2006,
              <http://www.rfc-editor.org/info/rfc4427>.

   [RFC5212]  Shiomoto, K., Papadimitriou, D., Le Roux, JL., Vigoureux,
              M., and D. Brungard, "Requirements for GMPLS-Based Multi-
              Region and Multi-Layer Networks (MRN/MLN)", RFC 5212,
              DOI 10.17487/RFC5212, July 2008,
              <http://www.rfc-editor.org/info/rfc5212>.

   [RFC5521]  Oki, E., Takeda, T., and A. Farrel, "Extensions to the
              Path Computation Element Communication Protocol (PCEP) for
              Route Exclusions", RFC 5521, DOI 10.17487/RFC5521, April
              2009, <http://www.rfc-editor.org/info/rfc5521>.

   [RFC5557]  Lee, Y., Le Roux, JL., King, D., and E. Oki, "Path
              Computation Element Communication Protocol (PCEP)
              Requirements and Protocol Extensions in Support of Global
              Concurrent Optimization", RFC 5557, DOI 10.17487/RFC5557,
              July 2009, <http://www.rfc-editor.org/info/rfc5557>.

   [RFC5623]  Oki, E., Takeda, T., Le Roux, JL., and A. Farrel,
              "Framework for PCE-Based Inter-Layer MPLS and GMPLS
              Traffic Engineering", RFC 5623, DOI 10.17487/RFC5623,
              September 2009, <http://www.rfc-editor.org/info/rfc5623>.

   [RFC5671]  Yasukawa, S. and A. Farrel, Ed., "Applicability of the
              Path Computation Element (PCE) to Point-to-Multipoint
              (P2MP) MPLS and GMPLS Traffic Engineering (TE)", RFC 5671,
              DOI 10.17487/RFC5671, October 2009,
              <http://www.rfc-editor.org/info/rfc5671>.

   [RFC6163]  Lee, Y., Ed., Bernstein, G., Ed., and W. Imajuku,
              "Framework for GMPLS and Path Computation Element (PCE)
              Control of Wavelength Switched Optical Networks (WSONs)",
              RFC 6163, DOI 10.17487/RFC6163, April 2011,
              <http://www.rfc-editor.org/info/rfc6163>.





Zhang & Minei                 Informational                    [Page 21]


RFC 8051            Applicability for a Stateful PCE        January 2017


   [RFC7580]  Zhang, F., Lee, Y., Han, J., Bernstein, G., and Y. Xu,
              "OSPF-TE Extensions for General Network Element
              Constraints", RFC 7580, DOI 10.17487/RFC7580, June 2015,
              <http://www.rfc-editor.org/info/rfc7580>.

   [RFC7688]  Lee, Y., Ed. and G. Bernstein, Ed., "GMPLS OSPF
              Enhancement for Signal and Network Element Compatibility
              for Wavelength Switched Optical Networks", RFC 7688,
              DOI 10.17487/RFC7688, November 2015,
              <http://www.rfc-editor.org/info/rfc7688>.

   [RFC7698]  Gonzalez de Dios, O., Ed., Casellas, R., Ed., Zhang, F.,
              Fu, X., Ceccarelli, D., and I. Hussain, "Framework and
              Requirements for GMPLS-Based Control of Flexi-Grid Dense
              Wavelength Division Multiplexing (DWDM) Networks",
              RFC 7698, DOI 10.17487/RFC7698, November 2015,
              <http://www.rfc-editor.org/info/rfc7698>.

Acknowledgements

   We would like to thank Cyril Margaria, Adrian Farrel, JP Vasseur, and
   Ravi Torvi for the useful comments and discussions.

Contributors

   The following people all contributed significantly to this document
   and are listed below in alphabetical order:

   Ramon Casellas
   CTTC - Centre Tecnologic de Telecomunicacions de Catalunya
   Av.  Carl Friedrich Gauss n7
   Castelldefels, Barcelona 08860
   Spain
   Email: ramon.casellas@cttc.es

   Edward Crabbe
   Email: edward.crabbe@gmail.com

   Dhruv Dhody
   Huawei Technology
   Leela Palace
   Bangalore, Karnataka 560008
   India
   Email: dhruv.dhody@huawei.com







Zhang & Minei                 Informational                    [Page 22]


RFC 8051            Applicability for a Stateful PCE        January 2017


   Oscar Gonzalez de Dios
   Telefonica Investigacion y Desarrollo
   Emilio Vargas 6
   Madrid, 28045
   Spain
   Phone: +34 913374013
   Email: ogondio@tid.es

   Young Lee
   Huawei
   1700 Alma Drive, Suite 100
   Plano, TX 75075
   United States of America
   Phone: +1 972 509 5599 x2240
   Fax: +1 469 229 5397
   Email: leeyoung@huawei.com

   Jan Medved
   Cisco Systems, Inc.
   170 West Tasman Dr.
   San Jose, CA 95134
   United States of America
   Email: jmedved@cisco.com

   Robert Varga
   Pantheon Technologies LLC
   Mlynske Nivy 56
   Bratislava 821 05
   Slovakia
   Email: robert.varga@pantheon.sk

   Fatai Zhang
   Huawei Technologies
   F3-5-B R&D Center, Huawei Base
   Bantian, Longgang District
   Shenzhen 518129
   China
   Phone: +86-755-28972912
   Email: zhangfatai@huawei.com

   Xiaobing Zi










Zhang & Minei                 Informational                    [Page 23]


RFC 8051            Applicability for a Stateful PCE        January 2017


Authors' Addresses

   Xian Zhang (editor)
   Huawei Technologies
   F3-5-B R&D Center
   Huawei Industrial Base
   Bantian, Longgang District
   Shenzhen, Guangdong  518129
   China

   Email: zhang.xian@huawei.com


   Ina Minei (editor)
   Google, Inc.
   1600 Amphitheatre Parkway
   Mountain View, CA  94043
   United States of America

   Email: inaminei@google.com































Zhang & Minei                 Informational                    [Page 24]