scispace - formally typeset
Search or ask a question
Proceedings ArticleDOI

PCE: What is it, how does it work and what are its limitations?

TL;DR: An overview of three PCE deployment models in the software defined network (SDN) control architecture and the formal decoupling of the path computation allows more flexibility in the deployment of PCEs in other control paradigms outside their original scope.
Abstract: We overview the PCE architecture and how it can mitigate some weaknesses of GMPLS-controlled optical networks. We identify some of its own limitations and the way they are being addressed, along with its deployment models in SDN/Openflow.

Summary (4 min read)

Introduction

  • In a routing and distributed resource assignment approach, the source node only computes the spatial path using TED information (e.g., link aggregated unreserved bandwidth), but resource assignment is performed at the destination and intermediate nodes during signaling in the backwards direction.
  • First, this paper overviews in Sec.II, the Path Computation Element (PCE) architecture to overcome the source-based routing shortcomings.

II. BASICS OF THE PCE ARCHITECTURE

  • The main idea is to decouple the path computation function from the GMPLS controllers into centralized and dedicated PCE with an open and well-defined interface and protocol (Fig.2.a).
  • In the distributed path computation model(Fig. 3.b), several PCEs can be deployed, each one serving requests from a subset of GMPLS controllers, and where computation of paths is shared among the PCEs.
  • After the initial handshake, a PCC can request point-to-point or point-to-multipoint path computations, by sending a path computation request to the PCE (PCReq message) with a variety of objects that specify the set of constraints and attributes.
  • The PCReq message specifies the endpoints (source and destination node addresses) and objective functions (requested algorithm / optimization criteria), and the associated constraints such as traffic parameters (e.g. requested bandwidth), the switching capability, and the encoding type.
  • Such a mechanism could be either incremental (like IGP) or involving a bulk transfer of the complete TED.

III. LIMITATIONS OF GMPLS-CONTROLLED OPTICAL NETWORKS AND PCE-BASED SOLUTIONS

  • To this end, optical layer physical impairments may need to be taken into account by the path computation, in addition to the network topology and resources, in order to provide a path with adequate QoT, which renders the problem even more complex.
  • The distributed routing approach requires extensions to the 5 standard GMPLS OSPF-TE routing protocol to disseminate information on physical impairments, besides existing TE attributes.
  • As described in [15], there are two components contributing to the OSNR level estimation: the Link OSNR that considers the total ASE noise induced by all the (pre- and line) amplifier spans of a link and the Node OSNR which considers the ASE noise caused by the node booster amplifier.
  • On the other hand, the distributed signalling approach requires extensions to the GMPLS RSVPTE to gather information on physical impairments parameters, in order to encompass the required QoT whilst establishing the optical connections [16].

B. PCE-based solution for impairment-aware path computation

  • The PCE allows to overcome the limitations of impairmentaware distributed control plane.
  • Thus, impairment information is neither collected nor flooded by the GMPLS controllers.
  • Then, the PCE can use both the TED and the ETED for the path computation.
  • The system consists of a main multithreaded asynchronous process, running as a PCEP/TCP server in order to accept and process path computation requests from the PCCs.
  • It is based on a distance-adaptive, iterative, two-phase approach, combining dynamic spatial path computation (i.e., strict list of nodes an links) based on the TED, and off-line path characterization assuming Coherent Optical Orthogonal Frequency-Division Multiplexing transmission stored in the ETED, used for dynamic modulation and spectrum assignment, depending on requested bitrate and the physical distance.

C. Multi-domain path computation problem

  • Multi-domain networks are becoming a key component in core transport architectures, given the fact that an operator’s network may include several equipment vendors and segmenting the network into domains (as Interior Gateway Protocol -IGP - areas or Autonomous Systems -AS) is a means to increase the overall scalability and/or for confidentiality reasons.
  • The source node determines the next domain and the ingress within that domain.
  • The main limitation of the distributed per-domain path computation is that the egress domain nodes (or Area Border Routers - ABR) determined by the source node may be not optimal, due to the loss of global topology visibility.
  • Let’s consider the example of Fig.6.a to illustrate the problem.
  • Based on the TED of domain 1, Node A computes an ERO composed of node A,C and D (all strict hops, since they are within the same domain) and Q (as loose hop), since it is the optimal path within Domain A.

D. PCE-based solutions for multi-domain path computation

  • Most initial research efforts on the PCE-based multi-domain path computation were targeted to improve or extend the backwards recursive path computation (BRPC) procedure [18].
  • Each PCE is responsible for the path computation within its domain.
  • Standard BRPC procedure does not consider wavelength continuity constraint in WSON.
  • Then, the source domain PCE may send the computed i-VSPT to the downstream PCE to compute its own i-VSPT (the shortest paths from the source node to all domain egress nodes).
  • Finally, the parent combines the domain segments for an end to end path and sends the reply to originating child PCE / PCC.

E. Multi-layer path computation problem

  • Multi-layer transport network integrating both packet and optical circuit switching technologies (e.g., packet switching capable - PSC for MPLS-TP, and lambda switching capable - LSC for WSON), leverages the high-bandwidth transport capacity and deterministic performance provided by the optical circuit switched technology, as well as the efficient traffic aggregation and statistical multiplexing provided by packet switched networks.
  • To fully exploit the advantages provided by the traffic grooming it is necessary to promote the cooperation among the layers by means of a peer control plane interworking model.
  • In a multi-layer network, a LSP must initiate and terminate at the same specific layer and may traverse one or more lower layers.
  • Let’s consider the example of Fig.10 to illustrate the computation and provisioning of a multi-layer path with a GMPLS unified control plane with PCE.
  • Most of the FA TE link attributes are inherited from the associated LSC FA LSP.

F. PCE-based solutions for multi-layer path computation

  • In [24], the first implementation and lab trial demonstration of a unified control plane for multi-layer (MPLS-TP/WSON) networks was presented.
  • In other words, the routing and signalling protocols of each control plane layer act independently.
  • In general, multi-layer TE relies on both an optimal multilayer path computation and the automated provisioning of all involved layers.
  • An entity named Virtual Network Topology Manager (VNTM) is able to coordinate the layered establishment of server and client connections, also known as Layered provisioning.
  • First, the NMS requests the computation of a multi-layer path between H1 and H4 to the PCE located in the higher layer (e.g., packet).

A. TED synchronization

  • As described in Sec.II, a PCE operates with network state information (topology and resources) collected in the TED provided by a synchronization mechanism.
  • Thus, a PCE computes paths based uniquely on TED information which may not be synchronized with the actual network state, leading to an increase in the blocking of the connections during the provisioning.
  • This mechanism allows synchronizing all the nodes’s TED repositories within a given time after a network state change, referred to as the routing convergence time (in this example the authors assume whenever there is a change in the network state, a link update state is disseminated just after the change).
  • This may result on simultaneous path computation requests if multiple LSPs were interrupted, that may reduce the restorability.
  • That is, a limited form of statefulness is applied, known as contextawareness [6], [29].

B. Lack of global LSP state

  • The lack of global LSP state information (e.g., LSP route and reserved resources) may result in sub-optimal PCE algorithms.
  • The non-linear effects of optical fibers such as Cross-talk (XT), Cross-Phase Modulation (XPM) or Four-wave mixing (FWM) generate adjacent wavelength/ frequency slot interference that may cause the provisioning of a new optical connection to degrade the quality of service (QoS) of the in-service connections.
  • Thus, impairment-aware RMSA algorithms should compute new paths that ensure an acceptable QoS of the existing ones.
  • Since, there is no path with enough available bandwidth, the algorithm must preempt some LSPs to free some bandwidth.
  • If the PCE would have computed the requested path with global LSP state, it would have selected the path (LSR1,LSR4,LSR5,LSR6,LSR7), that despite having one additional link, it only requires to preempt one LSP.

C. Control of path reservations

  • The PCE lack of control of path reservations (e.g modification or rerouting) may also result in sub-optimal PCE algorithms.
  • Thus, LSP2 would not fail if the PCE could re-route LSP1 through LSR1-LSR4-LSR5-LSR6.
  • To this end, an active stateful PCE allows for optimal path computation considering the LSPDB, but not only as an input of the path computation process, but also for the control of the state (e.g. rerouting) of the stored LSPs.
  • Moreover, an active stateful PCE can also request the modification/rerouting of the existing LSPs (e.g., to reoptimize the path) to the source GMPLS controllers (headend PCC of the connection).
  • Fig.18.b provides an example of a LSP created and later removed by the PCE.

VI. PCE DEPLOYMENT MODELS IN SOFTWARE DEFINED NETWORKS

  • Software-defined networks (SDN) are based on three key pillars, as shown in Fig.19.
  • This controller maintains a detailed, global, updated and unique view of the network state (TED) and connection state .
  • This single controller is responsible for computing paths in a centralized way and for provisioning the connections in a centralized vertical manner (i.e., without requiring horizontal coordination among the nodes).
  • In brief, OpenFlow allows to identify and abstract the common set of functions from switches (both packet and circuit) and routers, and offer them in an standardized way to the centralized controller [36].
  • Finally, if the PCE is an active stateful PCE with instantiation capabilities, the PCE becomes a full controller (provisioning, modification and release of LSPs) and therefore, it could be integrated with the SDN/OpenFlow controller (Fig.20.c).

VII. CONCLUSIONS

  • The authors have overviewed the PCE architecture and how it can mitigate some weaknesses of GMPLS-controlled WSON/SSON.
  • This paper has identified two main trends.
  • Thus, the stateful PCE would require synchronization of the LSPDB and coordination of the path computation by communicating with each other.
  • Third, path computations considering both TED and LSP databases would be highly complex, requiring large amounts of computer processing resources.
  • Thus, the particular integration of the PCEs within the SDN/OpenFlow centralized control model becomes an opportunity, since it would allow reuse of the know-how of algorithms developed in the scope of PCE in the OpenFlow controller.

Did you find this useful? Give us your feedback

Content maybe subject to copyright    Report

Citations
More filters
Journal ArticleDOI
TL;DR: Monitor information from an operating network combined with supervised machine learning (ML) techniques is used to understand the network conditions and propose two supervised ML regression models, implemented with Support Vector Machine Regression (SVMR), to estimate the individual penalties of the two effects and then a combined model.
Abstract: For reliable and efficient network planning and operation, accurate estimation of Quality of Transmission (QoT) before establishing or reconfiguring the connection is necessary. In optical networks, a design margin is generally included in a QoT estimation tool (Qtool) to account for modeling and parameter inaccuracies, ensuring the acceptable performance. In this article, we use monitoring information from an operating network combined with supervised machine learning (ML) techniques to understand the network conditions. In particular, we model the penalties generated due to i) Erbium Doped Fiber Amplifier (EDFA) gain ripple effect, and ii) filter spectral shape uncertainties at Reconfigurable Optical Add and Drop Multiplexer (ROADM) nodes. Enhancing the Qtool with the proposed ML regression models yields estimates for new or reconfigured connections that account for these two effects, resulting in more accurate QoT estimation and a reduced design margin. We initially propose two supervised ML regression models, implemented with Support Vector Machine Regression (SVMR), to estimate the individual penalties of the two effects and then a combined model. On Deutsche Telekom (DT) network topology with 12 nodes and 40 bidirectional links, we achieve a design margin reduction of ∼1 dB for new connection requests.

51 citations

Journal ArticleDOI
TL;DR: This paper presents the first prototype implementation and experimental evaluation of an active stateful PCE with instantiation capabilities for the GMPLS-controlled flexi-grid DWDM network of the ADRENALINE testbed.
Abstract: Adaptive flexi-grid optical networks should be able to autonomously decide where and when to dynamically setup, reoptimize, and release elastic optical connections, in reaction to network state changes. A stateful path computation element (PCE) is a key element for the introduction of dynamics and adaptation in generalized multiprotocol label switching (GMPLS)-based distributed control plane for flexi-grid DWDM networks (e.g., global concurrent reoptimization, defragmentation, or elastic inverse-multiplexing), as well as for enabling the standardized deployment of the GMPLS control plane in the software defined network control architecture. First, this paper provides an overview of passive and active stateful PCE architectures for GMPLS-enabled flexi-grid DWDM networks. A passive stateful PCE allows for improved path computation considering not only the network state (TED) but also the global connection state label switched paths database (LSPDB), in comparison with a (stateless) PCE. However, it does not have direct control (modification, rerouting) of path reservations stored in the LSPDB. The lack of control of these label switched paths (LSPs) may result in the suboptimal performance. To this end, an active stateful PCE allows for optimal path computation considering the LSPDB for the control of the state (e.g., increase of LSP bandwidth, LSP rerouting) of the stored LSPs. More recently, an active stateful PCE architecture has also been proposed that exposes the capability of setting up and releasing new LSPs. It is known as active stateful PCE with instantiation capabilities. This paper presents the first prototype implementation and experimental evaluation of an active stateful PCE with instantiation capabilities for the GMPLS-controlled flexi-grid DWDM network of the ADRENALINE testbed.

23 citations

Journal ArticleDOI
TL;DR: This paper proposes and experimentally demonstrates a technique based on a periodically poled lithium niobate waveguide to achieve both frequency conversion and defragmentation in elastic (or flex-grid) optical networks.
Abstract: Super-channel (or multi-carrier) transmission is today one of the most promising techniques for the support of high line rates, which are required to satisfy the massive increase of Internet traffic. Moreover, flex-grid optical networks seem to be the candidates for backbone networks by enabling high spectral efficiency thanks to the adoption of the ITUT flex-grid. Such networks may suffer from spectrum fragmentation, which can prevent the establishment of new connections. For this reason, defragmentation techniques (i.e., reoptimization) have been widely studied, especially considering single-carrier transmission. Inparallel, the software defined networking (SDN) paradigm and the active stateful path computation element (PCE) are emerging as candidates for the control of next-generation optical networks. Such architectures are also particularly suitable in the case of defragmentation since they enable the controller to trigger reoptimization procedures. In this paper, we investigate defragmentation in the presence of super-channels, at both the control and data planes. We propose and experimentally demonstrate a technique based on a periodically poled lithium niobate waveguide to achieve both frequency conversion and defragmentation in elastic (or flex-grid) optical networks. Its peculiarity is that it is suitable for super-channels because it avoids detrimental subcarrier overlapping during a frequency shift. SDN with the OpenFlow protocol is discussed for the control of such operations, as well as the active stateful PCE and generalized multi-protocol label switching (GMPLS). The frequency conversion and defragmentation techniques are demonstrated in a lab trial considering a 200 Gb/s super-channel and extended OpenFlow for the control plane. No loss of data is experienced.

22 citations


Cites background from "PCE: What is it, how does it work a..."

  • ...Spectrum fragmentation may also be overcome with reoptimization techniques, i.e., defragmentation, consisting of reallocating connections (e.g., changing their frequency position) such that new requests may find enough contiguous spectrum to be served [18–21]....

    [...]

Journal ArticleDOI
TL;DR: Several use cases where the active stateful architecture can provide some benefits are presented and discussed, including impairment-aware path computations in the context of multirate optical networks, recovery solutions, global defragmentation, and dynamic LSP adaptations.
Abstract: The path computation element (PCE) architecture was originally proposed with a stateless condition, i.e., considering only network reserved resources during constraint-based path computations. More recently, a stateful architecture was introduced to additionally maintain the state of computed and established label switch paths (LSPs). Furthermore, the PCE architecture evolved to active functionality, enabling the PCE to directly issue recommendations to the network. In this study, we present and discuss several use cases where the active stateful architecture can provide some benefits. They include impairmentaware path computations in the context of multirate optical networks, recovery solutions, global defragmentation, and dynamic LSP adaptations. The latter use case is then specifically demonstrated in a network testbed including a flexigrid optical network operated with a multicarrier 1 Tb/s transmission with coherent detection. Novel advanced digital signal processing (DSP) monitoring functionalities are introduced and experimentally demonstrated. These monitoring functionalities are utilized to trigger a new hitless dynamic adaptation technique operating on the applied low-density parity check (LDPC) transmitted coding. The technique has been successfully demonstrated to increase transmission robustness upon impairment degradation, such that no traffic disruption is experienced. Moreover, to accommodate the LSP coding adaptation, network reconfiguration has been performed, successfully driven by the PCE thanks to the active functionality.

18 citations


Additional excerpts

  • ...I. INTRODUCTION T he path computation element (PCE) architecture hasbeen proposed to enable effective constraint-based path computations [1–3]....

    [...]

Journal ArticleDOI
TL;DR: This work proposes several approaches that help find a pair of disjoint end-to-end paths that may traverse multiple domains from source to destination and result in minimum total cost.
Abstract: In a network in which multiple domains are defined due to geographical and/or administrative reasons, only a limited amount of domain information is exchanged by domain service providers. Topology aggregation is a method used to facilitate this limited information exchange. The amount of information provided for each domain may vary based on the technical and management decisions taken by the service provider. For instance, some domains may choose to provide only a single shortest path between two border nodes, while another may be able to provide a pair of disjoint paths with minimum total cost. In such cases, end-to-end protected path routing needs to facilitate and use different amounts of domain information provided by domain service providers in order to find the best solution. In this work, we propose several approaches that help find a pair of disjoint end-to-end paths that may traverse multiple domains from source to destination and result in minimum total cost. These approaches include methods for inter-domain information exchange that carry costs of disjoint paths within a domain. The performance of minimizing the total cost of a pair of end-to-end paths is investigated. Finally, the blocking probabilities of these various approaches due to the existence of trap topologies in the network are also discussed.

12 citations


Cites background from "PCE: What is it, how does it work a..."

  • ...Various types and levels of information can then be stored at path computation elements (PCEs) [10–12], which are responsible for path computation by communicating with each other to exchange required information....

    [...]

References
More filters
Journal ArticleDOI
31 Mar 2008
TL;DR: This whitepaper proposes OpenFlow: a way for researchers to run experimental protocols in the networks they use every day, based on an Ethernet switch, with an internal flow-table, and a standardized interface to add and remove flow entries.
Abstract: This whitepaper proposes OpenFlow: a way for researchers to run experimental protocols in the networks they use every day. OpenFlow is based on an Ethernet switch, with an internal flow-table, and a standardized interface to add and remove flow entries. Our goal is to encourage networking vendors to add OpenFlow to their switch products for deployment in college campus backbones and wiring closets. We believe that OpenFlow is a pragmatic compromise: on one hand, it allows researchers to run experiments on heterogeneous switches in a uniform way at line-rate and with high port-density; while on the other hand, vendors do not need to expose the internal workings of their switches. In addition to allowing researchers to evaluate their ideas in real-world traffic settings, OpenFlow could serve as a useful campus component in proposed large-scale testbeds like GENI. Two buildings at Stanford University will soon run OpenFlow networks, using commercial Ethernet switches and routers. We will work to encourage deployment at other schools; and We encourage you to consider deploying OpenFlow in your university network too

9,138 citations


"PCE: What is it, how does it work a..." refers methods in this paper

  • ...Second, a standardized interface between the controller and the data plane, for example based on the OpenFlow protocol [35] for connection provisioning and network discovery....

    [...]

01 Mar 2009
TL;DR: PCEP is designed to be flexible and extensible so as to easily allow for the addition of further messages and objects, should further requirements be expressed in the future.
Abstract: This document specifies the Path Computation Element (PCE) Communication Protocol (PCEP) for communications between a Path Computation Client (PCC) and a PCE, or between two PCEs. Such interactions include path computation requests and path computation replies as well as notifications of specific states related to the use of a PCE in the context of Multiprotocol Label Switching (MPLS) and Generalized MPLS (GMPLS) Traffic Engineering. PCEP is designed to be flexible and extensible so as to easily allow for the addition of further messages and objects, should further requirements be expressed in the future. [STANDARDS-TRACK]

471 citations

01 Oct 2005
TL;DR: This document specifies a link management protocol (LMP) that runs between a pair of nodes and is used to manage TE links and will be used to maintain control channel connectivity, verify the physical connectivity of the data links, correlate the link property information, suppress downstream alarms, and localize link failures for protection/restoration purposes in multiple kinds of networks.
Abstract: For scalability purposes, multiple data links can be combined to form a single traffic engineering (TE) link. Furthermore, the management of TE links is not restricted to in-band messaging, but instead can be done using out-of-band techniques. This document specifies a link management protocol (LMP) that runs between a pair of nodes and is used to manage TE links. Specifically, LMP will be used to maintain control channel connectivity, verify the physical connectivity of the data links, correlate the link property information, suppress downstream alarms, and localize link failures for protection/restoration purposes in multiple kinds of networks. [STANDARDS-TRACK]

275 citations


"PCE: What is it, how does it work a..." refers background in this paper

  • ..., LMP [5]) for control channel connectivity maintenance, link property correlation, and fault localization....

    [...]

  • ...The GMPLS architecture defines a set of standard protocols constituted by three pillars: a signaling protocol (i.e., RSVPTE [3]) used for setting up the end-to-end connections (label switched paths— LSPs), a routing protocol (i.e., OSPF-TE [4]) used for topology and network resource dissemination, and a link management protocol (i.e., LMP [5]) for control channel connectivity maintenance, link property correlation, and fault localization....

    [...]

01 Apr 2009
TL;DR: This document specifies a procedure relying on the use of multiple Path Computation Elements (PCEs) in order to compute such inter-domain shortest constrained paths along a determined sequence of domains, using a backward recursive path computation technique while preserving confidentiality across domains, which is sometimes required when domains are managed by different Service Providers.
Abstract: The ability to compute shortest constrained Traffic Engineering (TE) Label Switched Paths (LSPs) in Multiprotocol Label Switching (MPLS) and Generalized MPLS (GMPLS) networks across multiple domains (where a domain is referred to as a collection of network elements within a common sphere of address management or path computational responsibility such as IGP areas and Autonomous Systems) has been identified as a key requirement . This document specifies a procedure relying on the use of multiple Path Computation Elements (PCEs) in order to compute such inter-domain shortest constrained paths along a determined sequence of domains, using a backward recursive path computation technique while preserving confidentiality across domains, which is sometimes required when domains are managed by different Service Providers.

250 citations


"PCE: What is it, how does it work a..." refers methods in this paper

  • ...path computation were targeted to improve or extend the backward recursive path computation (BRPC) procedure [18]....

    [...]

Journal ArticleDOI
TL;DR: Two distributed approaches that integrate information about most relevant physical impairments in RWA and lightpath provisioning are presented and assessed and functional requirements, architectural functional blocks, and protocol extensions for implementing either an impairment-aware real-time RWA, or a light Path provisioning based on impairment- aware signaling are provided.
Abstract: The absence of electrical regenerators in transparent WDM networks significantly contributes to reduce the overall network cost. In transparent WDM networks, a proper resource allocation requires that the presence of physical impairments in routing and wavelength assignment (RWA) and lightpath provisioning be taken into account. In this article a centralized, a hybrid centralized-distributed and two distributed approaches that integrate information about most relevant physical impairments in RWA and lightpath provisioning are presented and assessed. Both centralized and hybrid approaches perform a centralized path computation at the management-plane level, utilizing physical impairment information, while the lightpath provisioning is done by the management plane or the control plane, respectively. The distributed approaches fall entirely within the scope of the ASON/GMPLS control plane. For these two approaches, we provide functional requirements, architectural functional blocks, and protocol extensions for implementing either an impairment-aware real-time RWA, or a lightpath provisioning based on impairment-aware signaling.

141 citations


"PCE: What is it, how does it work a..." refers background in this paper

  • ...An impairment-aware distributed control plane can be classified as distributed routing or distributed signaling [13]....

    [...]

Frequently Asked Questions (1)
Q1. What have the authors contributed in "Pce: what is it, how does it work and what are its limitations?" ?

First, the authors present an overview of the PCE architecture and its communication protocol ( PCEP ). Then, the authors present in detail the considered source-routing shortcomings in GMPLS-controlled networks, namely, impairment-aware path computation, multi-domain path computation and multi-layer path computation, as well as the different PCE-based solutions that have been proposed to overcome each one of these problems. In this sense, the authors provide an overview of three PCE deployment models in the Software Defined Network ( SDN ) control architecture.