Zhang Yongjun,Yao Yu, Gu Wanyi
(Key Laboratory of Optical Communication and Lightwave Technologies, Ministry of Education, Beijing University of Posts and Telecommunications,Beijing 100876, P. R. China)
Abstract:The fast development of IP services forces the traffic flow in the networks to move g radually from Time Division Multip lexing(TDM)services to packet data services.To adap t to the change,low-cost Ethernet technologies are introduced into the Metropolitan Area Network(MAN).Ring network control p rotocols and ring network p rotection technologies are app lied to guarantee manageability and reliability required in the telecom networks.The Ethernet Ring Protec tion(ERP)and Resilient Packet Ring(RPR)are two d ifferent ring p rotection technolog ies.The Transport Multi-Protocol Label Sw itching(T-MPLS)network uses the T-MPLS Shared Protection Ring(TM-SPRing)technology to p rovide ring p rotection.Carriers can expand the p rotection as needed and use app rop riate ring technolog ies to meet the requirem ents of Carrier Ethernet(CE).
T he types ofservices carried by the telecom network have changed rad ically along w ith the acceleration of the IP service development.The p rimary service traffic is changed from Time Division Multip lexing(TDM)to packet.Because of the change ofservice type,there is a new requirement for the bearernetwork.The existing Synchronous DigitalHierarchy(SDH)/Synchronous Op ticalNetwork(SONET)-based transportnetwork is suitab le for constant-rate synchronization circuitservices,butit is notquite suitab le for the packetservices w ith burstand uncertainty features,especially the large g ranularity data services(GE or10GE).Therefore,the p rob lem is how to selecta suitab le packetbearernetwork technology to adap t to the tendency of IP traffic and p rovide the data service with availab le,effective and manageab le service,which is the same as that for the trad itional circuitservice.This is a g reatchallenge thatcarriers have to face in network construc tion[1].
To ensure the com petitive power for the Metropolitan Area Network(MAN),all the carriers are considering how to op tim ize or reconstruc t the existing networks tomeet the requirementof the new generation b roadband subscribers.The b roadband access,based on the Ethernet technology,w ith the feature of low-costand the support forhigh bandw id th,has become a p rimary access solution for the existing MANs.
The Ethernet technology is originally designed for the LocalArea Network(LAN)use,w ith no consideration forany extension to MAN orWide Area Network(WAN).But its ease-of-use and low cost make itw idely app licab le forb roadband access.Therefore,allthe world standard ization organizations set to research whether the telecom-c lass features can be introduced to the existing Ethernetand thus imp lement end-to-end Ethernetservices.
The so-called Carrier Ethernet(CE)[2]is to realizemanageability and reliability in the Ethernetby add ing transport network functions,expand ing frame headeror introducing Layer2 p rotocols and signaling based on the frame structure of the trad itionalEthernet.As a MAN Ethernet technology,Ethernet ring system solves the p rob lem s of the trad itionaldata network such as low p rotection and slow fault resiliency.
Theoretically,itcan p rovide a feature of 50ms fastp rotection.Meanwhile,it is compatib le w ith trad itionalEthernet p rotocols.Therefore,itis an im portant technicalselec tion and solution forMAN network op tim ization and reconstruction.
Ethernetuses Carrier Sense Multip le Accessw ith Collision Detect(CSMA/CD).An Ethernet ring may cause b roadcast storm,whichmakes the entire network fullofdata traffic and results in network b reakdown.A tree structure is generally used in Layer2 network construction because of this technical feature.Each node has only one unique up link node,so the network w illnotbecome a ring,resulting in the fact thatEthernethas no ring p rotectionmechanism.Thismakes it very difficult to achieve telecom-c lass reliability and availability.Ethernet ring technologies are generated based on the concep tofSDH ring p rotec tion.The technicalp rincip le is to sw itch over to the standby link quickly using ring network topology upon device or link failure,w ith no affection on the running services.In the telecom field,the 50ms levelof fault resiliency canmeet the requirementof service continuity.
ITU-TG.8032 p rovides a specific purpose and definition of the Ethernet ring networking[3-4].Itdefines the Ethernet ring as a ring topology composed ofa group of Ethernetnodes compatib le w ith IEEE 802.1.Each node is connected to the other two nodes through the 802.3 Med ia Access Control(MAC)-based ring port.EthernetMAC can be carried by otherservice-layer technologies(such as SDHVC and MPLSEthernet pseudow ire).Allthe nodes can communicate w ith each otherd irectly or ind irec tly.Meanwhile,an Ethernet ring is both a physicaltopology and a logic topology,in which traffic flow is forwarded fully based on the forwarding rules of the IEEE 802.1 specification.It supports Ethernetservices such as point-to-point(E-Line),multipoint-to-multipoint(E-LAN),and point-to-multipoint(E-Tree)services,inc lud ing EthernetPrivate Line(EPL)and EthernetVirtualPrivate Line(EVPL).It supports various communicationsmodes inc lud ing unicast,multicastand broadcast,and can p reventdata disorderand dup lication.Ethernet ring supports threemodes of ring interconnections:shared nodes,a link composed of two shared nodes,and multi-ring/layernetwork constructed by overlapped Ethernetnetworks.Services can be transported from end to end through interconnected rings.
Itis often seen in actualapp lications that ord inary Ethernetsw itches are used to estab lish a ring-topology network,such as triangularand square topologies.However,these topologies are not ring technology.In simp le words,ring network is the basis and p rerequisite of the ring technology.Besides physical ring topology,the devices using the ring technology also supports ring control p rotocols.Forexamp le,when sw itches are used to estab lish a square network,the network is only a ring topologically,butgenerally itdoes notuse any ring technology or p rotocol.The devices'forwarding and otheroperations are not d ifferent from themesh network,so such networking is nota ring technology.Only the ring-topology network using a ring controlp rotocolis a ring technology.
▲Figure 1. ERP ring architecture.
There are two c lasses ofCE ring p rotec tion technologies:EthernetRing Protec tion(ERP)and ResilientPacket Ring(RPR).All the existing Ethernet ring technologiesmeetcarriers'requirements fornetwork performance through relevant technology expansion.
The ERP technology is a link layer p rotocolfor Ethernet ring only.Itp revents data ring from generating broadcast storms in the Ethernet ring.When a link or device on the Ethernet ring fails,itcan be sw itched over to the standby link quickly to ensure fastservice recovery.The fast fault resiliency p rincip le of the ERP p rotocolis to transfer ring control messages through the p rivate controlling VirtualLocalArea Network(VLAN)inside the ring.Meanwhile,w ith the topological features of the ring itself,itd iscovers network failures quickly and sw itches the active link over to the standby one for fast recovery[5-6].
Different from otherp rotec tion mechanisms,ERP p rovides p rotec tion for multicastservices w ithout the need of allocating additionalcapacity on the ring.All the capacities on the ring can be used to carry the p rotec ted multicast traffic.
According to the currentEthernet ring p rotec tion sw itching p rotocol(G.8032/Y.1344),the ERP technology needs to configure themain controlnode in each ring before working,w ith other nodes as transportnodes.Themain controlnode can detect rings quickly by send ing and receiving relevantp rotocol frames.According to the shortestdelay p rincip le,itselects a ring port from a node on the ring and sets it to the b locked status[7].Node D in the ERP ring architecture dep icted in Figure 1 is set to the"Blocked"status,while the ports of othernodes on the ring are set to the forward ing status.The b locked port p rocesses only relevantp rotocol messages thatuse the Lightweight Distributed EthernetProtection Algorithm(LDEPA)and d iscards otherdata and controlmessages,thus d isconnec ting the ring logically and solving the p rob lem ofbroadcaststorm on the ring.
When a link on the ring fails,the transm ission sw itch detects the link failure and immed iately sends a message to the p rimary sw itch.The p rimary sw itch unb locks the b locked port immediately to rap id ly recoverdata forward ing.Othernodes on the ring c lear the forward ing add ress tab le when they receive a failure notificationmessage.If the node has a b locked porton the ring,itsets the port to the forward ing status.
When the failure is recovered,the node thatd iscovers the failure recovery confirms the recovery through the waiting recovery time,and then notifies a failure recoverymessage to the two sides of the ring,and keeps the portconnec ted to the failed span to the b locked status.
The RPR technology is a MAC layer p rotocoldefined in IEEE 802.17 for op tim izing data service transm ission in a ring architecture[8].Itcan adap t to multip le physicallayers(such as SDH,Waveleng th Division Multip lexing,and Ethernet)and transm itmultip le types of services effectively,such as voice,data and video.RPR defines a brand new frame structure[9],which is a two-layer MAC add ress.RPR frame allows RPR to be a ring p rotocoland technology by adding special fields such as Time-to-Live(TTL)and Control-related ring control fields.The RPRMAC p rotocolis also d ifferent from the ordinary Ethernet.Besides using externalMAC add resses(for identifying ring node devices),RPR sw itching also uses TTL,Controland other fields to select rings and check frame TTL.RPR also defines an independentand sophisticated Operation,Adm inistration and Maintenance(OAM)frame and p rotocol for the RPR technology only,which are used for topology d iscovery,fault detection,performance detec tion,and fairness control.
RPR has a bid irectionaldual-ring topology.Each segmentofop tical channelon the ring works at the same rate.The d ifference is thatboth rings of the RPR can transm itdata.The two rings are called Ring let0 and Ring let1.Ring let0 transm its data in the c lockw ise d irec tion and Ring let1 transm its data in the counter-c lockw ise d irection.
Each RPR node uses a 48-bitMAC add ress,which is used in the Ethernet,as an add ress ID.Therefore,the two pairs ofsending/receiving physical op ticalinterfaces are only a link layer interface as viewed from the link layerof the RPR node devices.As viewed from the network layer,only one interface IP add ress is required.The link between two ad jacentRPR nodes is called"span".Many continuous spans and nodes on the spans form a domain[10].RPR uses an SDH ring architec ture and has a very high capability of fault resiliency,which allows faultp rotection sw itching w ithin 50 ms.When a link failure occurs,Ring let0 and Ring let1 are connected to each other inside the nodes on both ends of the failed link to form a new ring.
The above two ring p rotec tion technologies are mainly based on the trad itional Ethernet.They ensure the telecom-c lassmanageability and reliability of the Ethernet.
The Multi-ProtocolLabel Sw itching(MPLS)technology allows sw itching based on label.Italso supports data service transm ission.TransportMulti-Protocol LabelSw itching(T-MPLS)is a unified multi-service transm ission technology thatmeets telecom-c lass requirementbased on the inherited MPLS data features.T-MPLS based ring network is also called T-MPLSShared Protection Ring(TM-SPRing)[11].T-MPLS ring p rotection is a technology thatp rovides fastservice p rotection in TM-SPRing.
T-MPLS ring is a T-MPLS LSP transm ission ring constructed through logic struc turemapp ing.Allthe nodes in a TM-SPRing estab lish logicalad jacent relationships.Connection relationships are estab lished w ithoutany restric tion of physicaldevices orMAC topology.The connection between two ad jacentnodes is called span.In a T-MPLS ring,span is a bidirectionalconnec tion thatcan be both physicaland logical.The transport path entity for transporting services between ring nodes is realized by a g roup of T-MPLS based LabelSw itching Paths(LSPs).The ring bearerentity is in a dual-ring struc ture formed by one or more LSPs.The service traffic on the two rings runs in opposite direc tions,c lockw ise and counter-c lockw ise.Accord ing to the specific service direction,they can be divided into working ring(working service direction)and p rotection ring(opposite to the working service d irec tion).Multip le LSPs can be estab lished on each ring according to the numberofservices.Different LSPs are allocated for d ifferent service flow.TM-SPRing p rovides p rotection for the span between ad jacent nodes.Service p rotection sw itching depends on the quality ofsignals in the span.Figure 2 illustrates the T-MPLS ring architecture.
▲Figure 2. T-MPLS ring architecture.
To p reventinvalid span between ad jacentnodes,a p rotec tionmechanism should be determ ined to p rovide fast p rotection in the case of failure.At p resent,two p rotectionmechanism s are used in TM-SPRing:steering and w rapping.The steering mechanism allows the source node to change the path when a failure is detected.It bypasses the failed node and transm its data flow directly to the destination node.The w rapp ing mechanism allows the node near the failure to loop the data flow back to another ring.Itallows the data flow tomaintain the connection to the destination node by using a long path.The greatestd ifference between the two mechanism s is that the nodes that red irectdata flow upon failure are d ifferent.The steering mechanism allows the source node to send data flow.The w rapp ingmechanism allows the ad jacent node near the failure to red irect the data flow.The w rapp ing p rotection sw itch requires a shortstartup time,w ith fewer packets discarded,but the p rotected path to which itsw itches is not the op timalroute.The p rotected path that uses the steering mechanism is the op timal route,butit requires a long switching startup time,w ithmore packets discarded.
In T-MPLSSPRing,the Automatic Protection Sw itch(APS)message achieves faultdetection and p rotection switching by referring to the fast p rotection sw itching technology used in SDH.In normalconditions,a node w ithoutany fault reports its normalstatus to ad jacentnodes,which are destination nodes.In case ofa failure,the node that discovers the failure originates an app rop riate b ridging requestin both directions.The purpose is to transm it information between the firstand the end nodes sw itched in the w rapp ing mode,and to transm it the information of the failed span in the g lobalnetwork.The source node of the ASP information is the node thatdiscovers the failure,and the destination node is its ad jacentnode,i.e.,the othernode of the failed span.Allthe non-destination nodes transm itAPS information transparently.In the case when the failed span is invalid,the APS information between the source and destination nodes w illpass through all the nodes on the ring.Each node p rocesses the APS information and takes a sw itching ac tion to achieve service p rotection on the ring.
The InternationalTelecommunication Union Telecommunication Standard ization Sec tor(ITU-T)is responsib le for the T-MPLS ring p rotection scheme.Itmainly uses the SDH functionmodule p rotocolG.783 as a reference,but the standard ization is not com p lete,and it requires further normalization ofsome issues(such as the source routingmode and mesh network expansion).
In the MAN,ring architecture ismainly used for the transportnetwork.Ring network topology hasmany advantages in layered networking,saving line resources(e.g.,op tical fiberand cab le),p roviding fastand flexib le p rotec tion,and simp lifying network topology and networkmanagement.These advantages are notavailab le in line and star networks.However,itmustp rovide data channels w ith fast fault resiliency capabilities,in which themost im portant one is the p rotection sw itching w ithin less than 50m s.
Atp resent,in the networks thatare using various CE technologies,mostof the vendors'devices support ring networking[12].These technologies are largely sim ilarwithm inor d ifferences.They are ring network topologies physically and chain or tree topologies logically.The numberofnodes on the ring is related to the device performance,traffic flow and transm ission d istance,but there is no restriction theoretically.Most of the existing ring technology solutions are expanded from the RPR or ERP technology based on the ITU-T Recommendation G.8032,forexam p le,Multip le SuperRing(MSR),ZTEEthernet SmartRing(ZESR)and Ethernet-Shared Protec tion Ring(E-SPRing).Themain d ifferences inc lude d ifferent fault detec tion app roaches and the flexib le selec tion of congestion points,which can im p rove the faultdetec tion time and network bandw id th utilization.The TM-SPRing technology developed for T-MPLSw illgradually be adop ted by carriers along w ith thematuration and im p rovementof T-MPLS and the w ide dep loymentof T-MPLSnetworks.TM-SPRing w illbe used tomeet the requirementofCE reliability and manageability.
The firstp rob lem thatneeds to be solved forCE technology is high reliability.With the ring topology,CE allows fastand effective sw itch-over to the standby link w ithoutany im pac ton the services when a network device or link fails.The p rotec tion switchingmeets the requirementof50ms.Each Ethernet ring technology has its own em phasis.The ERPand RPR technologies are expanded accord ing to the app lication demands tomeetcarriers'requirements.
The TM-SPRing p rotection technology w illalso be w idely researched along w ith the deep research and w ide use of T-MPLS.Therefore,the usage and dep loymentof the technologies and devices ofd ifferentvendors w ith the consideration ofcarriers'p ractical requirements are themost im portant issues for p rovid ing ring network p rotection for the carriers.