Role of Network Management in Network Restoration and ...jakab/edu/litr/TMN/old/Restoration_Mngmt.pdf · role of network management in network restoration and resilience ... 1354

A R C H I T E C T S O F A N I N T E R N E T W O R L D

Role of NetworkManagement in NetworkRestoration and Resilience

T E C H N O L O G Y W H I T E P A P E R

The introduction of a distributed control plane in

the transmission nodes, improvements in the

path control mechanisms in packet switched

networks, and ever increasing pressure on

operating expenses are affecting the role

network management plays in network

restoration and resilience. Intelligent traffic

restoration mechanisms can reduce the (unused)

bandwidth reserved for protection and provide

an acceptable recovery time in the event of

transmission path problems. As this new

technology needs to be implemented in existing

networks, an evolutionary approach is required.

Consequently it is essential for the path control

functions between existing centralized network

management systems to be able to coexist with

the distributed control plane driven path

controls.

New network management system features will have an importantrole in managing the successful evolution of optics and data networksto implement distributed control plane architectures.

P. Fogliata, C. Wilson, M. Ragni

ROLE OF NETWORKMANAGEMENT IN NETWORKRESTORATION AND RESILIENCE

1 | Alcatel Telecommunications Review - 3rd Quarter 2003

IntroductionThe past few years have seen tremendous changes

for all players in the telecommunication arena. Ever-increasing demand from “users” for more and moreinnovation in the access, edge, metro and corenetworks includes:

• More bandwidth and added value, consisting ofservice diversity and control requirements.

• A revolution in terabit routing, switching andtransport infrastructures, such as Optical TransportNetworks (OTN).

• Innovative optical network elements, protocols, andmanagement and control systems used ininnovative network configurations.

Today, data services are the dominant users ofbandwidth in many service providers’ networks. Also,because of the dominance of enterprise applications,Internet Protocol (IP) traffic management and controlcapabilities are beginning to complement traditionalconnection-oriented traffic management systems. Inmany instances, service providers are also attemptingto differentiate their service offerings by consideringadvanced network architectures based on newintelligent transmission technology. Thesearchitectures focus on reducing operatingexpenditures and improving network managementflexibility by taking advantage of intelligent controlfunctions in the nodes.

New control capabilities are being introduced intonodes in both optical Time Division Multiplex (TDM)1

transmission networks and cell/packet switchednetworks. It is difficult to predict the pace of thistransition, which will take place at different times for

the TDM and data layers. Consequently, the networkmanagement design must evolve, complementing thenew control plane functions. At the same time, it musttake into account the network operator’s migrationscenario when the transition takes place in anoperational environment.

Control Plane ModelsDefinition of Control Plane

The control plane is implemented by deployingintelligence in the nodes, enabling them to gatherinformation about the network environment. Figure 1

shows the control plane as a logical layer, parallel tothe ‘data plane’. The real-time collection of networkconfiguration data enables network nodes tocollaborate, offering the possibility to performcoordinated reconfiguration actions without theintervention of an external manager.

1 Here, TDM is used as shorthand for Synchronous Digital Hierarchy (SDH),Synchronous Optical NETwork (SONET) and Dense Wavelength DivisionMultiplex (DWDM) optical transmission technologies.

Control Plane

Data PlaneUNI UNI

UNI

Fig. 1 Relationship between the control plane and data plane

UNI: User–Network Interface

A set of protocols between the control plane partsexchange network information and coordinate anyreconfiguration. Recently there has been a flurry ofactivity in several standards groups aimed at marryingMulti Protocol Label Switching (MPLS) andSDH/SONET/DWDM networking technologies to definea protocol set that is also suitable for a control plane inthe TDM world. In this case, MPLS is referred to asGeneralized MPLS (GMPLS). GMPLS extends thenotion of label swapping with MPLS to timeslots andwavelengths. The emergence of the Private Network toNetwork Interface (PNNI) routing protocol supports asimilar trend in Asynchronous Transfer Mode (ATM)networks.

The International Telecommunications Union –Telecommunications (ITU-T) Automatic SwitchedOptical Network (ASON) architecture gives anextended view, with the network being composed oflayered domains that interact with other domains in astandard way. The “user” is an endpoint device that canrequest connection services dynamically over a UNI.

Role of Network ManagementOne inherent feature of a control plane is the ability

to reroute traffic in the event of a fault, a featureknown as restoration. This mechanism requires all thenetwork resources to be permanently monitored,including those that are not used, to ensure that theyare available should rerouting be required.

The functionality assigned to the control plane mayvary, but the correct strategy is to use it to enhance thenodes with multi-vendor and real-time capabilities.Trying to exchange all the management informationusing the control plane will result in it being overloaded,

eventually leading to a critical performance problem.Thus the control plane will complement rather thanreplace the Network Management (NM) system, asshown in Figure 2. NM will have to ensure consistencybetween the network resources and the carrier’s internaldatabase. Other features related to faultreporting/handling, accounting, performance monitoringand security would remain under NM control, which willnow also include management of the control plane.

Control plane actions and network managementoperations are designed to be complementary. The roleof NM is to prepare the configuration data to enablethe control plane, acting as a real-time tool, to set upand tear down services and preserve them in the eventof network faults. Traffic engineering tools are onesuch application; they provide the right configurationparameters for optimal distributed routing decisions.Once correctly configured, the control plane will beable to operate even if the NM is unavailable. On theother hand, NM must be able to override the controlplane if necessary. For example, the status of resourcescould be changed and hidden from the control plane.The visibility of the network state is an essential NMfeature, enabling the operator to repair a faultycomponent (with the traffic already automaticallyrerouted) and to control the network so that itoperates in a stable and predictable manner.

Evolution of Network ManagementApplications

The evolution of NM applications must take intoaccount the benefits of the control plane functionsimplementing data path service routing, mechanisms torecognize services affected by failed resources, andreconfiguration of the data path across the controllednetwork. Two evolutionary aspects are consideredbelow:

• Evolution of the Alcatel 1354 Network Manager(NM) in optical networks with the introduction of adistributed control plane in the OTN.

• Alcatel 5620 NM improved path controlmechanisms in cell/packet switched networks.

Key Elements of Intelligent Optical NetworksThere are two scenarios for implementing an optical

network with intelligent nodes: a “greenfield”implementation or evolution from an existing networkin which the nodes have no control plane capabilities.The second scenario is analyzed here because it doesnot preclude the first one; it might even facilitate it.

Centralized control in OTNsIn this scenario, protection mechanisms like linear

Multiplexer Section Protection (MSP), Sub-Network

ROLE OF NETWORK MANAGEMENT IN NETWORK RESTORATION AND RESILIENCE

Alcatel Telecommunications Review - 3rd Quarter 2003 | 2

Control Plane

Data Plane

Management Plane

UNI UNI

UNI

Fig. 2 Adding the management plane

Connection Protection (SNCP), Drop & Continue(D&C), and Multiplexer Section Shared ProtectionRing (MS-SPRing) (two or four fibers) are used toincrease fault resilience in both the metro and coretransport networks. Protection mechanisms sometimesrequire an ad hoc communication protocol betweenthe various network elements involved. Protection isnormally set up and supervised by a central networkmanager which implements management functions likepath provisioning, network configuration, faultreporting/handling, accounting, performancemonitoring and security.

The Alcatel 1354 Regional Manager (RM) providesthis type of Network Management System (NMS) forboth SDH and (D)WDM networks. In the core of thenetwork it is common to have meshed subnetworkswhere a central NMS provides fault resilience through“restoration” actions, to improve bandwidthoptimization compared with the previously listedprotection architectures. This NMS collects real-timealarm information from the Network Elements (NE)and reconfigures the meshed subnetwork based on aglobally optimized view.

The Alcatel 1354 Network Protector (NP) is one ofthe most efficient restoration managers, capable ofrearranging an optical core subnetwork within a fewseconds, even in the event of multiple faults.

To provide a single point of management for thewhole network and to share large NEs betweenprotection- and restoration-based mechanisms, the1354 RM acts as a manager of managers, providing anintegrated view, including the subnetworks controlledby the 1354 NP. End-to-end path provisioning iscomputed by the 1354 RM, which requests a nominalroute across the restoration-based subnetworks to the1354 NP. In the same way, typical managementfunctions (path setup, fault management, etc) arehandled for the whole network, as shown in Figure 3.

Distributed control in OTNsBy adding more intelligence to the NEs in

restoration-based subnetworks, the previous scenariocan be implemented using distributed restorationcontrol within the subnetworks. This additionalintelligence is realized by adding the ability to takelocal routing decisions based on topology and resourceinformation which is exchanged with other NEs thatare members of the subnetwork.

In Alcatel NEs, the Generalized MPLS RoutingEngine (GMRE), a software package deployed in eachnode, optionally implements this distributed controlplane in existing nodes (see Figure 4). The add-oncapability of the control plane is complemented by theessential ability to migrate step-by-step (or to assignpart of) the node resources to a set controlled by thedistributed control plane instead of to the centralnetwork manager, as far as circuit provisioning controlis concerned.

As information sharing and connection setup ordersare performed through standard signaling protocols(GMPLS), this solution is open to a multi-vendorenvironment.

An additional advantage of this distributed approachis that restoration decisions are taken by each node, innear real-time, even if the NMS is not present or notconnected to the NE. The disadvantage is that eachnode takes its restoration decisions autonomously,without any coordination. Consequently, suchdecisions may not be globally optimized or may beslow to converge in large, highly loaded subnetworks.To help overcome this last drawback, better standardalgorithms are required which take into accountcontributions from the central network management topre-calculate recovery routes and strategies.

In any case, the central network manager retains itsvarious roles, including fault reporting/management,



Centralized Protection Network Manager(Alcatel 1354 RM)

Restoration Manager(Alcatel 1354 NP)

AccessNetwork

CoreNetwork

SNCP D&C Linear MS P 2 fiber MS-SPRing

4 fiber MS-SPRing

Meshed Net

Fig. 3 Centralized network management

Centralized Network Manager(Alcatel 1354 RM)

GMRE GMRE

GMREGMREGMRE

Fig. 4 Adding distributed control plane functionalityto OTN NEs

performance monitoring, and security. An additionalfeature of the 1354 RM enables the operator to ask theNM to activate the setup of circuits in order to sendcommands to the appropriate GMRE, which will initiatesignaling throughout the distributed controlledsubnetwork.

Evolution from existing OTNsNew intelligent networks hold out the promise of

reducing capital expenditure and operating expenses(through multi-vendor compatibility, advancedscalability, faster provisioning, network auto-discovery,efficient bandwidth usage for restoration etc) andproviding new services (like bandwidth on demand).

Moving from the traditional to the new types ofnetwork, Alcatel is proposing an evolutionaryapproach whereby established networks are “boosted”by central applications to provide similar functionalityto that of the new distributed control planetechnologies.

As an example, in the case of bandwidth on demand,a central GMPLS proxy “enriches” the established non-GMPLS networks to provide, in real-time, circuits inresponse to requests coming from a user through theUNI (so-called switched connections). In this way, thisfunction is available not only across the new GMPLS-capable network, but also through the traditionalnetwork.

In this case, the GMPLS proxy – Alcatel 1355Bandwidth on Demand (BonD) – reroutes the UNIrequests through the NMS that interfaces the non-GMPLS NEs(see Figure 5). The central networkmanager acts as a common repository in which theUNI request is verified for compliance with existingService Level Agreements (SLA).

A similar approach is used for restoration wheretraditional subnetworks supported by a centralrestoration manager coexist and interwork withnetworks based on GMPLS restoration .

On top of the complete network, the Alcatel 1354 RMprovides a uniform view and uniform procedures tominimize the impact on the operational process. Thisapproach also facilitates evolution where thedistributed control plane is not being introduced in agreenfield environment.

Key elements in 5620 managed networksThe Alcatel 5620 NMS has established its reputation

as a data network management system because of itsend-to-end provisioning capabilities, which aredelivered ubiquitously across the broad range ofequipment types and associated path types that itmanages (circuit-switched data, circuit-switched voice,ATM, frame relay, MPLS, Ethernet, etc).

In much the same way as in OTNs, the traditional NMparadigm is based on control from a centralmanagement plane. The emergence of the PNNIrouting protocol within the ATM Forum has providedan opportunity to realize the improvements offered bya distributed control plane. Similar techniques areemployed in MPLS networks to add traffic engineeringand Quality of Service (QoS) capabilities to IPnetworks.

PNNI or MPLS will typically be introduced inestablished networks, rather than in new or“greenfield” environments. ATM networks tend to be amixture of ATM and non-ATM equipment and will oftenevolve into a central PNNI network core with segmentsof non-PNNI capable equipment connected at thenetwork edge. Similar islands of MPLS capable and

non-MPLS capable NEs will alsobe typical of IP networks. Asophisticated NMS must providethe necessary measures to ensurea seamless approach to managingsuch networks.

PNNI/MPLS overviewThe enabling of PNNI in ATM

networks and MPLS in IPnetworks removes the tasks ofroute selection and connectionadministration from the NM anddistributes them to the sourceendpoint of each path or SoftPermanent Virtual Circuit (SPVC)or Label Switched Path (LSP).The PNNI/MPLS-enabledmultiservice switch (or “node”)that hosts the source endpoint forthe path, stores a prioritized list



Centralized Network Manager

CentralizedRestoration Manager

Restorationmanaged Network

Non GMPLSSubnetwork

GMPLSSubnetwork

DistributedRestoration Adapter

SLARepository

UNI

UNI UNI

GMRE

GMREGMRE

GMPLS Proxy

Fig. 5 Interworking between GMPLS and non-GMPLS subnetworks

of routing tables that define possible routes from itselfto each of the other “nodes” that it can see in thenetwork. These routing tables are generated via a peer-to-peer exchange of topology information with itsneighboring nodes. The improvement in connectionestablishment rates is substantial since each node inthe network can act unilaterally and asynchronously toconnect or restore the SPVCs under its control. Anadditional performance benefit is derived from the factthat route selection does not need to be computed atconnection time. This performance improvementcreates opportunities for network resilience andrecovery strategies that rely on the reroutingperformance of the network rather than on simplephysical line and/or card redundancy.

Evolution from the existing ATM to PNNI and IP to MPLS networksWhile a compelling argument can be made for service

providers to migrate existing ATM networks to PNNInetworks and IP to MPLS networks, such a migration isa logistical challenge. (How do we get there fromhere?) The act of PNNI enabling an ATM networkinvolves considerable effort in the areas of networkconfiguration and establishment of the requirednetwork infrastructure. The key components of a PNNInetwork are:

• Addressing scheme to uniquely identify the variousnetwork elements.

• Web or mesh of links supporting inter-nodal routingand signaling messages within the PNNI network.

The Alcatel 5620 supports the configuration of thenecessary point-to-point Virtual Path Connection Links(VPCL), each of which has an associated PNNI link(supporting the exchange of PNNI routing information,also called a Routing Control Channel or RCC) and asignaling link (used for sending connection requests).In addition to these basic network building blocks, thePNNI standard provides the ability to scale networks tolarge numbers of nodes by grouping them into ahierarchy, with each level represented by a single nodein the group – the Logical Group Node (LGN). This isachieved by applying a three-tiered hierarchy to thePNNI addressing scheme in which each node has aunique address at each level in the hierarchy, and asingle node represents the LGN as peer group leader.

As might be expected, a substantial amount ofmanual configuration is required to establish thesehierarchical networks. However, some information iscommon to the nodes in a given subtree within thehierarchy (i.e. summary address information). The5620 facilitates these configuration exercises byautomatically propagating any change to a single nodethat triggers the need for the same change to be madeto other nodes.

Similarly, in MPLS networks the IP links betweenrouters must be configured to carry MPLS signalingmessages. A good NMS solution, such as the 5620,would enable an operator to easily configure and viewthe MPLS-enabled parts of the IP topology, specify howmuch of the link bandwidth should be reserved forMPLS forwarding and IP forwarding, and which MPLSsignaling protocols are enabled on the link.

Once the requisite PNNI/MPLS networkinfrastructure (i.e. signaling and routing links) hasbeen established, a set of management functions is stillneeded to manage the ATM virtual channels or MPLSLSPs, even if the actual connection setup is performedin the network.

In the ATM case, there is usually a substantiallogistical hurdle in completing the network migration:how to convert the existing network managementsystem controlled PVCs to PNNI controlled SPVCs.Clearly, this conversion must be performed withminimal service disruption to the provider’s customers.A tool in the 5620 performs this PVC to SPVCconversion, connecting paths and verifying theestablishment of the new connection. Should one of theconverted paths fail to connect for any reason, the5620 automatically restores the original PVC.

When the PNNI network is fully operational, thechallenge for the NMS is to provide operators withsimilar levels of control and network visibility as werepreviously offered by the central management of the PVCnetwork. The 5620 provides the operator with a numberof tools to influence and correct (as needed) the routestaken by paths through the PNNI network. The OperatorDirected Routing (ODR) tool allows an operator tospecify a list of the NEs through which a given path mustpass. Similar LSP management functions are provided inthe MPLS case. The routes that an LSP may take throughthe network can be determined using the NMS point andclick capabilities on a view of the MPLS topology. Inaddition, MPLS realizes the concept of strict and looserouting; the NMS application allows the operator tospecify whether an LSP must use or may use aparticular router as it transits the network.

In addition to managing LSP routing, the 5620 canprovide numerous LSP configuration parameters toease the administration of MPLS networks (e.g. easysetting of traffic management parameters andclassification parameters).

In the ATM case, the Alcatel 5620 provides asophisticated console for administering network pathoptimization. Path optimization of SPVC segments isachieved using the ATM Forum standard “bridge androll” technique, known as “domain-based routing”.This technique allows the PNNI network to connectthe intermediate segment of the proposed new routefor an SPVC before releasing the established path.This dramatically reduces data loss, allowing



optimization to be achieved in less than 50 ms –effectively a “hitless” connection move.

Hybrid ATM and PNNI controlled ATM networksIn the ATM case, there are other scenarios in which

advanced management tools can make a difference tooperational excellence, even for networks withdistributed connection control planes. One suchexample is referred to as hybrid PVC and SPVCnetworks. This situation arises because firstly few“greenfield” data networks are deployed using ahomogeneous technological approach, and secondly,distributed routing protocols such as PNNI pertain to aparticular technology (i.e. ATM). While such atechnology may exist in one segment of a network, theservice provider’s end-to-end data path will probablytraverse or terminate on other legacy technologies.

In order to support this, the NM must know about theentry and exit points of the PNNI network. Figure 6shows how this is done with the 5620; the NMSprovisions two separate PVC segments in addition to anSPVC segment, resulting in a seamless end-to-endmanaged path commonly referred to as a hybrid SPVC(a concatenation of PVC and SPVC segments).

To provision such a path, a network operator needonly select the two endpoints for the path and specifythe desired path attributes (bandwidth, servicecategory, etc). The 5620 automatically identifies all thepath points required along the PVC segments, theentry and exit points of the PNNI network for theSPVC segment, and sends all the connection requestsneeded to complete the path. While the NMS is notinvolved in selecting and connecting the route for theSPVC segments, it does select the source anddestination endpoints for the SPVC and sends aconnection request to the selected SPVC source node.Upon receipt of this request, the SPVC source nodesignals a setup request, specifying a selected routefrom itself to the specified destination node.

Additional monitoring and diagnostic capabilities for PNNI/MPLS networks

In addition to end-to-end provisioning capabilities,network operators need to maintain the end-to-endvisibility of network paths. The strength of thiscapability directly influences the service provider’sability to provide a quality service. To support thisendeavor, the Alcatel 5620 provides a diagnosticinterface for the use of ATM Forum standardconnection trace and also for MPLS ping and trace

route operations. In the ATM case, this tool allowsusers to query the network and display a drawingshowing the network elements traversed by a givenSPVC. The ATM Forum path trace routine is usedwhen attempting to identify why a particular call setupfailed. The resulting drawing clearly shows the routesthat were used in an attempt to set up the call, and thereasons why the call failed to establish along thespecified routes. These utilities are integrated with theAlcatel 5620 route trace utility, which is used forsimilar diagnostic efforts on the PVC segments of ahybrid SPVC to maintain end-to-end visibility of thecomplete network path.

In the MPLS case, diagnostic tools are provided toenable operators to launch traditional IP ping andtrace route commands, direct which LSP the testpacket should use, and specify how many packets ofwhich size to send. The results can be displayedgraphically for easy viewing by the operator, showinginformation such as the route traveled by the packetand the delay experienced.

The need for network visibility extends beyond thevisualization of single end-to-end paths to thegathering of statistical data that can accuratelydescribe the performance and the behavior of thePNNI/MPLS signaling network. The reliability andperformance of the network carrying the signaling androuting messages is the foundation of the QoSdelivered by the network as a whole.

ConclusionThe move towards introducing network control

intelligence in the nodes requires an evolutionarynetwork management approach to support networkoperations, taking maximum advantage of thefunctional distribution between the control plane andthe network management applications.

Different technologies offer different solutions forsimilar problems, taking into account technology-specific issues. Furthermore, the introduction of theseinnovations should take into account interworking withexisting parts of the overall network. Alcatel offers acomprehensive set of network management systemsthat provide a consistent vision of the solutions neededto realize this evolutionary approach across differentnetwork technologies.



non-PNNIsubnetwork

PNNIsubnetwork

non-PNNIsubnetwork

Alcatel 5620 NMS

Fig. 6 Interworking between PNNI and non-PNNIenabled subnetworks



Paolo Fogliata is Director of Network ManagementR&D in the Alcatel Optical Networks Division,Vimercate, Italy. ([email protected])

Craig Wilson is Product Manager responsible forNetwork Management of ATM switching equipmentwithin Alcatel’s Fixed Networks Division, Ottawa,Canada. ([email protected])

Mario Ragni is responsible for the networkmanagement system team in the Alcatel OpticalNetworks Division, Vimercate, Italy.([email protected])

AbbreviationsASON Automatic Switched Optical Network

ATM Asynchronous Transfer ModeBonD Bandwidth on DemandD&C Drop & Continue

(D)WDM (Dense) Wavelength Division MultiplexingGMPLS Generalized Multi-Protocol Label Switching

GMRE GMPLS Routing EngineIP Internet Protocol

ITU-T International Telecommunications Union-Telecommunication

LGN Logical group NodeLSP Labelled Switched Path

MPLS Multi-Protocol Label SwitchingMSP Multiplex Section Protection

MS-SPRing Multiplex Section Shared Protection RingNE Network Element

NM(S) Network Management (System)NP Network Provider

ODR Operator Directed RoutingOTN Optical Transport Network

PNNI Private Network to Network InterfacePVC Permanent Virtual CircuitQoS Quality of ServiceRCC Routing Control ChannelRM Regional Manager

SDH Synchronous Digital HierarchySLA Service Level Agreement

SNCP Sub-Network Connection ProtectionSONET Synchronous Optical Network

SPVC Soft Permanent Virtual CircuitTDM Time Division MultiplexingUNI User-Network Interface

VPCL Virtual Path Connection Link



Alcatel and the Alcatel logo are registered trademarks of Alcatel. All other trademarks are the property of their respective owners. Alcatel assumes no responsibility for the accuracy of the information presented, which is subject to change without notice. © 01 2003 Alcatel. All rights reserved. 3GQ 00005 0010 TQZZA Ed.01

Documents

Role of Network Management in Network Restoration and ...jakab/edu/litr/TMN/old/Restoration_Mngmt.pdf · role of network management in network restoration and resilience ... 1354